aia-website/www/reports/migration-summary.md

# Migration Summary

## Status

A maintainable Astro migration and substantial fidelity-first pass are
implemented. The live public site was used as the authority for the shared
shell, English and Spanish homepages, service-page family, team page, library
indexes, and blog-post family.

## Migrated Content and Assets

- 46 English, 12 Spanish, and 7 Arabic library posts.
- 21 English, 9 Spanish, and 2 Arabic page records.
- One author record per language.
- 167 self-hosted images, 20 font files, and two PDF downloads.
- Current June and May 2026 English library articles are included.

HubSpot-generated wrappers, analytics, scripts, CSS, query-language duplicates,
AMP variants, pagination, and author archive variants are not carried into the
Astro implementation.

## Live-Source Fidelity Pass

Live pages inspected during the pass include:

- English and Spanish homepages
- ABA Therapy, Autism Evaluations, and Learner Social Club
- About, Team, Client Consultation, and Giveback & Donate
- English and Arabic Library indexes
- Representative current library articles

Implemented fidelity work includes:

- Rebuilt the live two-tier header, navigation hierarchy, language controls,
  utility links, footer columns, and contact information.
- Rebuilt the complete English and Spanish homepages in their live section
  order using one reusable Astro component and self-hosted source assets.
- Moved the English and Spanish homepage copy into section-oriented `home:`
  frontmatter blocks so the page is easier to edit in Front Matter CMS.
- Rebuilt service pages around the live compact title banner and editorial
  presentation, with current visible headings and working calls to action.
- Recovered and restored the live decorative hero imagery for all mapped
  English and Spanish page routes plus every language's library index.
- Restored the consultation form section's injected background image.
- Added a dedicated live-derived team card grid rather than presenting the
  extracted team content as a generic article.
- Rebuilt the contact page around the live two-column content/form section and
  full-width Leaflet map while retaining local clickable obfuscated phone
  links.
- Rebuilt the giveback page as four source-faithful full-width sections with
  its original image/copy proportions, fundraiser color band, nonprofit logo,
  emphasis, and external donation calls to action.
- Rebuilt library indexes and blog-post presentation around the live sidebar,
  article list, byline, featured-image, and counter patterns.
- Converted FAQ-bearing posts to MDX and routed their accordions, the
  standalone FAQ page, and the insurance FAQ through one reusable component.
  Each accordion uses the same animated, single-open interaction and a
  component-scoped caret color override; blog and insurance instances also
  emit matching `FAQPage` JSON-LD.
- Restored 32 live-source blog blockquotes with their original emphasis,
  links, fawn backgrounds, spacing, rounded presentation, and quote-mark SVG.
- Replaced generic oversized cards, rounded controls, and marketing heroes
  with live-derived typography, palette, widths, spacing, and compact controls.
- Corrected material live-source discrepancies found during the pass,
  including current homepage ESA copy, testimonial content, and ABA copy.

### Intentional visual deviations

- The homepage and `/insurance` carrier-logo rows intentionally share one
  responsive layout. The first and last logos sit flush with the row edges,
  and the Blue Cross Blue Shield mark is emphasized at 125% of the standard
  desktop logo width, following its treatment on the live homepage rather
  than preserving the live pages' differing row spacing.
- The Community 4 Autism logo on `/donate-autism-giveback` is horizontally
  centered on mobile instead of retaining the live page's left alignment.
- The empty `<h6>` retained by the live `/schedule-consultation` page was
  removed. It had no visible content and generated an empty heading ID in the
  migrated Markdown, so omitting it improves heading semantics and fragment
  integrity without removing page content.

## URLs and Redirects

Clean public URLs are preserved, including:

- `/aba-therapy`
- `/autism-evaluations`
- `/learner-social-club`
- `/client-consultation`
- `/library`
- `/library/post-slug`
- available `/es/...` and `/ar/library/...` routes

Phase 3A replaces the earlier 97-canonical sitemap with publication-policy
output: 92 local URLs in an indexing-enabled build and none in staging.
Nginx rewrites remain limited to the brief aliases `/aba`, `/autismevaluations`,
and `/learnersocialclub`.

## Forms

Visible consultation/contact-style pages render styled static HTML forms with
backend and spam-protection TODO comments. Submission is intentionally
disabled. Appointment and enrollment calls to action route to the static
consultation page rather than retaining the production Jotform backend.

The three English service pages also reproduce the live site's delayed,
right-edge inquiry drawers, including their page-specific enrollment copy,
shared field set, modal overlay, and responsive layout. These drawer forms are
static and their submit controls remain disabled until a production backend
and spam protection are selected; this is an intentional functional and visual
deviation from the live HubSpot form.

## External Dependencies

- The contact page map intentionally matches the live Leaflet/OpenStreetMap
  implementation and loads Leaflet from `unpkg.com` plus map tiles from
  `tile.openstreetmap.org`.

## Multilingual

- Spanish has the full live-derived homepage, translated service pages,
  library index, and available translated posts.
- Arabic has RTL support, a clearly marked placeholder landing page, the live
  library family, and available Arabic posts.
- Language-switcher choices are emitted only when the corresponding Astro
  route exists, preventing dead translated-route links.
- Arabic pages use `lang="ar"` and `dir="rtl"`; English and Spanish use LTR.
- No large Arabic translations were invented.

## Likes and Views

The post counter uses `PUBLIC_AIA_API_BASE`, defaulting to
`https://api.azinstitute4autism.com`. Its contract was verified against the
current live-site script:

- `POST /stats/batch` for likes and views
- `POST /likes` to like
- `DELETE /likes/{slug}` to unlike

It preserves cookie-backed liked state and degrades gracefully if the API is
unavailable.

## SEO and Accessibility

- Canonicals, Open Graph tags, Twitter card tags, semantic titles, and global
  organization JSON-LD are emitted by shared layouts.
- Organization JSON-LD intentionally omits raw `telephone` and `email` fields;
  visible contact details use the Astro obfuscation component to avoid exposing
  plain email and phone targets in rendered HTML.
- Library-index canonicals and current visible page H1s are explicitly set.
- Extracted image alt text is retained where available.
- Semantic landmarks, skip link, labeled forms, keyboard-operable navigation,
  language/direction attributes, and responsive layouts are present.
- Sätteri heading attributes remain enabled. Markdown supports explicit
  fragment IDs such as `{#accepted-insurance-heading}`; the installed MDX
  parser requires HTML heading IDs instead. The earlier claim that MDX also
  supports the shorthand was corrected during Phase 2 (see below).
- Sätteri directive parsing was disabled during Phase 2 after confirming that
  there are no intentional consumers and that it silently removes literal
  colon-containing text. No content or dependency versions were changed.

## Validation Results

- `npm audit --offline`: passed with zero known vulnerabilities under Node
  `24.18.0`; `flake.nix` provides a compatible current Node release.
- `npm run audit:links`: passed with zero broken internal source links. The
  audit now checks generated routes, root-relative references, relative
  Markdown links, and public assets when rendered output is unavailable.
- `npm run audit:images`: passed with no missing mapped page banners, section
  backgrounds, or blog featured images.
- `npm run audit:blog`: passed with no article-header content duplicated in
  blog content and valid MDX FAQ components for 125 questions across 19 posts.
- `npm run generate:sitemap`: passed; generated 97 URLs.
- `npm run generate:redirects`: passed.
- All migration `.mjs` tools and the sandbox DNS helper pass `node --check`.
- The Astro compiler parsed all 49 `.astro` files successfully.
- `npm run build`: passed with zero Astro diagnostics and generated 97 static
  pages.
- `npm run build:sandbox`: passed; generated 97 static pages.
- Live contact-page structure was inspected in browser at a 1280px desktop
  viewport; local generated HTML was checked for the contact layout, Leaflet
  assets, map container, and obfuscated clickable phone script. Local browser
  rendering was blocked because the Playwright browser cannot reach shell
  loopback servers and blocks `file:` URLs in this environment.
- Nix shell verification and `npm audit` retrieval were blocked by sandbox
  proxy/cache network resets.

## Manual Review

- Perform full-page desktop and mobile visual comparisons for page-specific
  layouts beyond the completed homepage, team, service, library, and article
  families.
- Editorially review extracted long-form content for remaining extraction artifacts,
  stale phone/email references, and heading hierarchy.
- Review all source alt text, keyboard behavior, screen-reader output, and
  contrast with accessibility tooling.
- Have fluent reviewers approve Spanish content and Arabic RTL presentation.
- Select and connect a form backend with spam protection before launch.
- Validate like/view CORS, credentials, and production behavior after deployment.

## Phase 2: Markdown Rendering Repair — 2026-10-07 UTC

These are newly verified local rendering findings. The accepted Phase 1 audit
and route evidence remain unchanged. No production requests were made, no
source content was rewritten, and no dependencies were upgraded.

### Inspection and Cause

Before changing configuration, searched the Markdown/MDX corpus, documentation,
parser configuration, scripts, plugins, layouts, and rendering components for
container, leaf, and inline directives and their consumers. There are no
intentional directives or registered directive transformations. Documentation
examples in code fences, ordinary punctuation, URLs, times, ratios, frontmatter
fields, CSS selectors, and explicit heading IDs are not directive consumers.

For all 100 content files, stripped frontmatter and inspected the installed
Sätteri `markdownToMdast` / `mdxToMdast` output with the original
`headingAttributes: true, directive: true` features. This found 24 unintended
`textDirective` nodes in seven files, with zero container or leaf directives.
Sätteri treats colon suffixes such as `:1`, `:00`, and `:Plan` as inline
directives; without a transformation, the renderer drops those nodes. Astro's
MDX integration inherits the same processor features, so both formats were
affected. The original full build and the failing regression build confirmed
the resulting text loss.

Only `directive` was changed to `false`; `headingAttributes: true` is retained.

| Source under `src/content/` | Restored literal text | Unintended nodes |
| --- | --- | ---: |
| `pages/en/aba-therapy.mdx` | `1:1`, `1:2` in headings and prose | 4 |
| `pages/es/aba-therapy.md` | `1:1`, `1:2` in headings and prose | 4 |
| `blog/en/aba-school-readiness-arizona.mdx` | `1:1`, `1:10` in prose and tables | 4 |
| `blog/es/aba-school-readiness-arizona.mdx` | `1:1`, `1:10` in prose and tables | 3 |
| `pages/en/learner-social-club.mdx` | `4:00 PM – 6:00 PM`, twice | 4 |
| `pages/es/learner-social-club.mdx` | `4:00 p. m. – 6:00 p. m.`, twice | 4 |
| `pages/en/faqs.mdx` | `Opportunities:Plan breaks` | 1 |

### Regression and Generated-Output Verification

`npm run test:markdown` creates a temporary Astro project, loads the real
`astro.config.mjs`, and uses the existing dependencies and copied source
components/content. It builds representative `.md` and `.mdx` fixtures plus
all seven affected collection entries through `astro:content`'s `getEntry` and
`render`. The fixtures never enter the production route tree, and all temporary
files are removed after the run.

The 20 tests cover `1:1`, `1:2`, `1:10`, ordinary spaced and adjacent colon
punctuation, English/Spanish times, plain URLs and link destinations (including
port/query/fragment), inline/fenced code examples, MDX embedded markup, explicit
heading IDs and fragment links, and the affected real content. Before the
configuration change, 14 tests failed and six passed. Afterward, all 20 passed.

The full build still generates 97 pages. Comparing the before/after `<main>`
HTML on all 97 pages found changes only in the seven routes corresponding to
the table above. This comparison normalizes the existing random
`ContactObfuscation` IDs in span attributes and script variables; that component
was not changed. Generated `/aba-therapy` and `/es/aba-therapy` each now contain
both ratios in their headings and prose (two occurrences of each ratio).

All existing explicit IDs remain intact. Four automatically generated service
heading IDs now reflect the restored text:

| Route | Previous generated ID | Corrected generated ID |
| --- | --- | --- |
| `/aba-therapy` | `aba-therapy-1-program` | `aba-therapy-11-program` |
| `/aba-therapy` | `aba-therapy-1-academic-readiness-program` | `aba-therapy-12-academic-readiness-program` |
| `/es/aba-therapy` | `programa-de-terapia-aba-1` | `programa-de-terapia-aba-11` |
| `/es/aba-therapy` | `terapia-aba-1-programa-de-preparación-académica` | `terapia-aba-12-programa-de-preparación-académica` |

No source references to the old IDs were found. Compatibility with production
fragment links remains part of the later URL reconciliation; this repair does
not introduce redirects or edit service content.

The installed MDX parser rejects Markdown's `{#id}` shorthand with directive
parsing either enabled or disabled. This is a pre-existing limitation, not a
regression from the repair. No current content uses that shorthand. Tests
verify Markdown shorthand and MDX's supported `<h2 id="...">` syntax. README
guidance was corrected; a broader parser change is outside this phase.

### Validation

Run from `www/`:

```sh
npm run test:markdown
npm run build
npm run audit:links
npm run audit:blog
git diff --check
```

- Markdown regression tests: 20 passed, zero failures.
- Build: passed; 74 checked files, zero errors/warnings/hints, 97 static pages.
- Link audit: zero broken internal links.
- Blog audit: zero content audit failures.
- Diff whitespace check: passed.
- No changes to application content, dependencies, lockfile, Phase 1 evidence,
  or `merge-plan.md`. No Phase 3 work was started.

## Phase 3A: Route Eligibility, Canonicals, and Indexing — 2026-10-07 UTC

This section records new implementation and verification. Earlier Phase 1
evidence and the Phase 2 checkpoint are preserved. No deployment, form
submission, backend activation, dependency change, content reconciliation,
hreflang, translation-alternate graph, article schema, or FAQ schema work was
performed.

### Publication Architecture

`src/utils/publication-policy.ts` is the shared, framework-independent
TypeScript policy. It derives local routes from collection/language/slug,
normalizes public paths, rejects route collisions (including drafts and utility
paths), validates canonicals, and decides route and sitemap eligibility.
`src/utils/publication.ts` adapts the Astro collections and reads the existing
`PUBLIC_ALLOW_INDEXING` switch. The canonical site origin comes from `site.json`,
also used by Astro configuration.

All content routes now pass through `src/pages/[...slug].astro`. The former
fixed home and Library routes could bypass `draft`; they now use the same
manifest as other pages. The existing Arabic home and three Library index
templates were moved to presentation components, retaining their visible
content. The English/Spanish home component receives its selected entry.
Shared layouts obtain canonical/indexing policy by the actual local route;
individual templates no longer invent or pass canonical strings.

The sitemap is an Astro endpoint generated alongside HTML. Its locations come
from eligible local route URLs, never canonical strings. The old public
sitemap copy was removed. `generate:sitemap` delegates to the Astro build
instead of maintaining a separate frontmatter-regex generator. Route discovery
and the source-link audit fallback reuse the shared publication policy; route
evidence now records eligibility, noindex, policy canonical, external-canonical
status, and production sitemap eligibility separately from rendered metadata.

The sole new frontmatter/schema field is `noindex`, a boolean defaulting to
false. Existing `draft` and `canonical` fields cover the remaining rules.

### Verified Exceptions and Read-Only Production Evidence

Six successful GET requests were made between **02:42:02 and 02:42:08 UTC**:
`robots.txt` first, then the five routes below. Requests were sequential, at
least one second apart, with redirects not followed. Robots allowed these
paths. No publisher pages, forms, or production APIs were requested. The web
tool could not retrieve robots.txt; the direct GET succeeded. All five route
observations agree with the accepted saved production evidence; these are
new checks, not silent revisions to historical findings.

| Route | Fresh HTTP / robots evidence | Implemented exception | Production sitemap |
| --- | --- | --- | --- |
| `/ar` | 404 at 02:42:08 | Existing placeholder marked `draft: true`; no HTML route | Excluded |
| `/schedule-consultation` | 200 / `noindex` at 02:42:04 | `noindex: true`; accessible, self-canonical | Excluded |
| `/employee-portal` | 200 / `noindex` at 02:42:05 | `noindex: true`; accessible, self-canonical | Excluded |
| `/library/community-highlight-meet-rula-diab` | 200 / `noindex` at 02:42:06 | `noindex: true`; external canonical retained | Excluded |
| `/library/new-aia-scottsdale-office` | 200 / `noindex` at 02:42:07 | `noindex: true`; external canonical retained | Excluded |

The exact verified external canonical targets are:

- [VoyagePhoenix original](https://voyagephoenix.com/interview/community-highlights-meet-rula-diab-of-arizona-institute-for-autism/)
- [Scottsdale.org original](https://www.scottsdale.org/airpark/features/integrity-empowerment-and-excellence-arizona-institute-for-autism-expands-to-new-office/article_79c31a18-2f35-11ee-9459-6f29f5a08420.html)

Only five content files changed, all in frontmatter: the Arabic home draft
flag, four noindex flags, and the two external canonical values. Portal access
and replacement behavior remain unchanged and deferred to human decision.

### Before / After and Output Verification

| Behavior | Before Phase 3A | After Phase 3A |
| --- | --- | --- |
| Route publication | Collection routes filtered drafts; fixed indexes bypassed them | Every content route uses the same eligibility manifest |
| Generated HTML | 97 pages, including `/ar` placeholder | 96 pages; only `/ar` removed |
| Canonical handling | Optional unchecked values passed by templates; syndicated posts self-canonicalized | Exactly one normalized self canonical, or the verified external exception; missing ordinary values derive from the route |
| Indexing-enabled robots | All pages defaulted to indexable | 92 `index,follow`; four verified `noindex,follow` exceptions |
| Staging robots | Global `noindex,nofollow` | Global protection retained on all 96 pages; no page override |
| Sitemap | Static list of 97 frontmatter canonicals | 92 eligible local URLs when indexing is enabled; empty in staging |
| Robots sitemap advertisement | Environment dependent | Still advertised only in the indexing-enabled build |

Published noindex pages remain accessible. An external canonical independently
excludes a page from the sitemap; it does not automatically invent a noindex
directive. The two real syndicated articles explicitly retain their verified
noindex status. Invalid/relative/unsafe canonicals, local targets that disagree
with the route or public origin, duplicate published canonical targets, and
route collisions fail validation. Duplicate frontmatter keys also fail parsing.

The language menu now uses eligible routes without a hardcoded `/ar` entry.
The Arabic header logo falls back to the existing English homepage while its
localized homepage is unpublished. No hreflang or new translation relationships
were introduced. Main content text, titles, and H1s are unchanged on all 96
retained routes. There are no stylesheet or asset changes.

The indexing-enabled local artifact was checked at **02:54:56 UTC**: all 96
heads matched their manifest canonicals/robots; 92 local sitemap URLs; no `/ar`.
The final ordinary build restores staging output with 96 noindex pages and
zero sitemap locations. No environment files or deployment configuration changed.

### Regression Coverage and Reproduction

`npm run test:publication` passes **19 tests** covering normal publication,
draft exclusion, accessible noindex pages, external canonicals independently
of noindex, normalized/missing/malformed/duplicate canonicals, cross-collection
and cross-locale collisions, reserved routes, staging protection, and all five
verified exceptions. Generated-head checks reject absent, duplicate, and invalid
canonical tags. It builds the actual application in temporary directories in
both environment modes, then additionally marks a home and Library index as
drafts and confirms their routes disappear. The tests never change real source,
environment configuration, or `dist`.

Validation from `www/`, in the requested order (with evidence refresh steps):

```sh
npm run test:publication
npm run build
PUBLIC_ALLOW_INDEXING=true npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
npm run audit:links
npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
git diff --check
```

All required commands passed. Each build checked 71 files with zero errors,
warnings, or hints, and generated 96 HTML pages. The link audit found zero
broken internal links. Existing regression suites also passed: 20 Markdown
tests and nine route-audit tests. Dependencies and the lockfile are unchanged.
The source-only link fallback also passed in an isolated directory with zero
broken links. The `generate:sitemap` compatibility command was verified using
Astro's public build API, followed by another normal staging build.

The final durable local reconciliation is
[`route-reconciliation-2026-10-07-offline-04-11-25-151Z.json`](route-reconciliation-2026-10-07-offline-04-11-25-151Z.json).
It reuses the earlier 70 production request records with their original
observation dates; neither offline reconciliation made new production requests.
Only the final staging reconciliation is retained in the repository; the
intermediate indexing reconciliation was moved to temporary storage after its
checks. The six targeted requests described above are separate fresh evidence.

The normalized discovery inventory remains 129 routes, with 102 URLs in the
saved production sitemap. Local generated routes decrease from 97 to 96;
sitemap/local overlap stays 92, and the ten sitemap-listed live-only routes
remain outstanding. `/ar` remains visible in the inventory as draft,
ineligible, not generated, and verified absent on production. Earlier audit
files retain their historical facts.

No new human publication-policy decision is needed for Phase 3A. Human gates
for portal access, forms/backends, analytics/consent, landing-page ownership,
and conflicting service facts remain deferred. Phase 3B was not started.

## Phase 3B: Explicit Translation Relationships and Hreflang — 2026-10-07 UTC

This section records new work after the accepted Phase 3A structural review.
Earlier audit findings and implementation history above remain historical.
Starting branch: `faithful-astro-migration`; HEAD: `1e1aafa`. The only commit
after Phase 3A (`cd097d0`) added the conversation transcript. The Phase 1 and
Phase 2 checkpoints remain `60c534d` and `ea2f1ea`. This work is not committed.

### Inspection and Policy

The existing manifest owns every generated content route and already supplies
eligibility, normalized local URL, canonical, noindex, external-canonical and
sitemap eligibility. Inspection found 32 page records, 65 blog records and
three author records, all declaring `translationKey`. No hreflang was emitted.
The language switcher was the only equivalent-route guesser: it removed a
language prefix and tried the same URL suffix in all languages. The Header's
locale-home fallback is navigation, not a translation relationship, and remains
unchanged.

`createPublicationManifest` in `src/utils/publication-policy.ts` now groups
explicit keys within each collection. It attaches one validated `translations`
array to every eligible member of a multilingual set. Each reference contains
the declared language and the target's actual manifest route and absolute URL.
It never derives an equivalent from a slug, filename or URL suffix.

Participation reuses `sitemapEligible`: a generated/published entry that is
indexable and self-canonical under the production policy. At least two eligible
languages are required. All members, including drafts, are checked for duplicate
language declarations before filtering; two entries for the same collection,
key and language fail validation. The graph validator rejects dangling targets,
wrong collection/key/language/URL, ineligible endpoints, duplicate languages,
missing self references and nonreciprocity. All members share the same complete,
ordered set, so reciprocity is generated rather than manually maintained.

**Grouping versus missing references:** `translationKey` is an optional grouping
label, not a pointer or a list of required locales. A singleton remains valid
and emits no alternates; it does not assert that a Spanish/Arabic record exists.
An absent referenced graph node fails policy validation; an absent referenced
HTML file fails rendered-output validation. A typo that creates a new singleton
or deletion of the last counterpart cannot be distinguished from intentionally
standalone content by this schema. No new required-target field or guessed
language requirement was introduced.

The existing optional `translationKey` schema now rejects empty and
whitespace-padded values in pages, blog and authors. There are no new frontmatter
fields and **no content/frontmatter edits**. Placeholders must remain drafts;
the only marked content placeholder is `/ar`, already `draft: true`.

### Current Explicit Sets

The existing declarations produce **24 sets / 53 participating routes**:
nine page sets (19 routes) and 15 blog sets (34 routes). English contributes
24 members, Spanish 21, Arabic eight. There are 19 two-language sets and five
three-language sets. The following inventory lists actual local routes, not
inferred URLs. A dash means there is no eligible declared counterpart.

| Collection / key | English | Spanish | Arabic |
| --- | --- | --- | --- |
| pages / index | `/` | `/es` | — |
| pages / library | `/library` | `/es/library` | `/ar/library` |
| pages / aba-therapy | `/aba-therapy` | `/es/aba-therapy` | — |
| pages / aba-therapy-intake-process | `/aba-therapy-intake-process` | `/es/aba-therapy-intake-process` | — |
| pages / autism-evaluations | `/autism-evaluations` | `/es/autism-evaluations` | — |
| pages / client-consultation | `/client-consultation` | `/es/client-consultation` | — |
| pages / contact | `/contact` | `/es/contact` | — |
| pages / learner-social-club | `/learner-social-club` | `/es/learner-social-club` | — |
| pages / services | `/services` | `/es/services` | — |
| blog / aba-school-readiness-arizona | `/library/aba-school-readiness-arizona` | `/es/library/aba-school-readiness-arizona` | — |
| blog / aba-school-readiness-guide | `/library/aba-school-readiness-guide` | — | `/ar/library/aba-school-readiness-guide` |
| blog / autism-evaluation-diagnosis-arizona-parent-guide | `/library/autism-evaluation-diagnosis-arizona-parent-guide` | `/es/library/autism-evaluation-diagnosis-arizona-parent-guide` | — |
| blog / autism-evaluation-what-to-expect | `/library/autism-evaluation-what-to-expect` | `/es/library/autism-evaluation-what-to-expect` | — |
| blog / autism-family-self-care-tips | `/library/autism-family-self-care-tips` | `/es/library/autism-family-self-care-tips` | — |
| blog / autism-self-advocacy-skills-aba | `/library/autism-self-advocacy-skills-aba` | `/es/library/autism-self-advocacy-skills-aba` | `/ar/library/autism-self-advocacy-skills-aba` |
| blog / behavior-management-functions-guide | `/library/behavior-management-functions-guide` | — | `/ar/library/behavior-management-functions-guide` |
| blog / emotional-regulation-aba | `/library/emotional-regulation-aba` | `/es/library/emotional-regulation-aba` | — |
| blog / enhancing-generalization-skills | `/library/enhancing-generalization-skills` | `/es/library/enhancing-generalization-skills` | — |
| blog / executive-functioning-skills-autism | `/library/executive-functioning-skills-autism` | `/es/library/executive-functioning-skills-autism` | `/ar/library/executive-functioning-skills-autism` |
| blog / first-then-cards-autism-transitions | `/library/first-then-cards-autism-transitions` | `/es/library/first-then-cards-autism-transitions` | `/ar/library/first-then-cards-autism-transitions` |
| blog / parents-guide-to-autism-and-aba | `/library/parents-guide-to-autism-and-aba` | `/es/library/parents-guide-to-autism-and-aba` | — |
| blog / positive-reinforcement-techniques | `/library/positive-reinforcement-techniques` | `/es/library/positive-reinforcement-techniques` | `/ar/library/positive-reinforcement-techniques` |
| blog / proactive-reactive-aba-strategies-guide | `/library/proactive-reactive-aba-strategies-guide` | — | `/ar/library/proactive-reactive-aba-strategies-guide` |
| blog / social-pragmatic-communication-autism | `/library/social-pragmatic-communication-autism` | `/es/library/social-pragmatic-communication-autism` | — |

Excluded: `/ar` is draft and uses `home`, not the published home set's `index`
key; its key was not changed. `/schedule-consultation`, `/employee-portal`,
`/library/community-highlight-meet-rula-diab`, and
`/library/new-aia-scottsdale-office` remain noindex; the latter two also retain
external canonicals. None can be a source or target. Three `authors:rula-diab`
records share a key but have no generated author routes and do not enter the
graph. The other 39 indexable routes have no eligible translated counterpart.
Live-only Spanish privacy, author/archive/search routes and missing languages
are not manufactured. There are no additional current URL-matched published
pairs lacking explicit keys; different-slug and unkeyed same-slug fixtures test
that future content will not be guessed.

### SEO, Navigation and Default Language

`Seo.astro` calls the shared `hreflangLinksFor` helper. An indexing-enabled build
emits absolute alternate URLs, including the page itself, on each set member.
English is the project's established default; `x-default` points to that set's
eligible English member, never an unrelated homepage. A set without eligible
English omits `x-default` (covered with Spanish/Arabic fixtures and excluded
English fixtures). All 24 current sets have an eligible English member.

This follows [Google's localized-page guidance](https://developers.google.com/search/docs/specialty/international/localized-versions)
on fully qualified, reciprocal, self-referencing alternates and an appropriate
default. Staging emits no hreflang; its global `noindex,nofollow` and empty
sitemap remain intact. The graph still supplies local language navigation.

`LanguageSwitcher.astro` uses those same manifest equivalents and the entry's
declared language for its active label. It constructs no counterpart URLs.
Labels, SVG and CSS are unchanged. Without a translation set, it retains the
current-language label but emits no empty menu, focusable trigger or listbox
promise. Existing translated sets keep self and equivalent links. Header and
route/template dispatch code are unchanged.

### Fresh Production Observations

Six read-only GETs on **2026-10-07 23:00:32–23:00:38 UTC**, at least one second
apart, checked robots first and then these five pages. Robots permitted the
paths; all six responses were HTTP 200, with no redirects followed. No forms,
APIs, broad crawl or production writes were used; no production HTML is added
to the repository. These checks supplement, rather than rewrite, earlier audit
evidence.

| Observed page | Check time UTC | Declared production alternates |
| --- | --- | --- |
| `/` | 23:00:33 | en `/`, es `/es/`, x-default `/` |
| `/library` | 23:00:35 | en-us `/library`, es `/es/library`, ar `/ar/library` |
| `/library/autism-self-advocacy-skills-aba` | 23:00:36 | en-us, es, ar, x-default English |
| `/es/library/autism-self-advocacy-skills-aba` | 23:00:37 | en-us, es, ar, x-default English |
| `/ar/library/autism-self-advocacy-skills-aba` | 23:00:38 | en-us, es, ar, x-default English |

The article's three observations reference the same three full article paths
listed in the set inventory. The homepage and article use English defaults.
Production's Library index has no x-default; the local policy adds its real
English index as the default, consistent with the approved rule. Library
production labels English `en-us`; local output uses the existing content
language `en`, without inventing regional variants. Relationships are based on
explicit existing content declarations, supported by these targeted checks;
this is not a new full content reconciliation or linguistic-fluency approval.

### Validation and Structural Review

`npm run test:publication`: **33 passed**, including the original 19 Phase 3A
tests and 14 translation tests. Coverage includes different-slug equivalence,
unrelated matching slugs, collection boundaries, all publication exclusions,
malformed/duplicate keys, dangling graph/generated references, reciprocity,
self references, default eligibility and identical switcher/SEO membership.
Isolated real-Astro builds exercise staging, indexing and draft home/Library
indexes without changing the real content or deploying anything.

The requested validation sequence passed (offline refreshes keep source
fingerprints current without making discovery requests):

```sh
cd www
npm run test:publication
npm run build
PUBLIC_ALLOW_INDEXING=true npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
npm run audit:links
npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
git diff --check
```

All three regular build invocations checked 71 files with zero errors, warnings
or hints, and generated 96 HTML routes. The link audit found zero broken links.
At 23:02:40 UTC, the indexing artifact had 92 sitemap entries, 92 indexable
pages and four noindex exceptions, with 174 hreflang links on 53 routes across
24 sets. Canonicals matched the pre-change artifact on every retained route.
At 23:03:49 UTC, the final staging artifact had 96 `noindex,nofollow` pages,
zero sitemap entries and zero hreflang links. Every language menu matched the
graph. Route set, titles, canonical URLs, language/direction and main content
matched the pre-change artifact; the main-content comparison excludes script
nodes and normalizes whitespace and existing random contact-obfuscation IDs.

The final durable route evidence is
[`route-reconciliation-2026-10-07-offline-23-03-24-581Z.json`](route-reconciliation-2026-10-07-offline-23-03-24-581Z.json).
It preserves the original 70 production request records and observation times.
The discovery inventory remains 129 routes, production sitemap inventory 102,
local generated 96, local/sitemap overlap 92 and live-sitemap local omissions
10. Those production counts are historical discovery evidence, not the local
92-entry production-policy sitemap. The intermediate indexing reconciliation
is retained only in temporary storage; both modes are reproducible above.

The diff is confined to the shared publication policy, translation-key schema
validation, switcher, SEO component, existing publication tests, README, this
report and the final dated route reconciliation. No content/frontmatter, route
templates, styles, dependencies, environment files or deployment settings were
changed. No Phase 3A blocker or accounting defect was found. The original
`merge-plan.md` remains untracked with SHA-256
`015db80cdbaf7d68799265d2070db760155c342b432ce6795f17ebc2218c641c`.

No new human publication-policy decision is required. Existing decisions about
employee access, campaign ownership, forms, analytics/consent and conflicting
eligibility claims remain deferred. Substantive future translations require
human language review. Phase 4A, article/FAQ schema, content reconciliation,
visual fidelity and deployment work were not started. Stop for review before
any Phase 3B checkpoint commit.

## Phase 4A — Display Headings and Social Metadata (2026-10-08)

This section records new implementation and observations after checkpoint
`32112edbde7f90d01becf5d420ff189bf8b7e6ec` on `faithful-astro-migration`.
Earlier audit and phase sections remain historical evidence. Phase 4B,
substantive content reconciliation, Library features and deployment are outside
this milestone. No changes have been staged or committed; stop for acceptance
of the implementation and structural review.

### Inspection and Architecture

Before changes, articles used `title` for both document title and H1. Regular
pages used `serviceTitles`, `pageTitles`, then the part of `title` preceding
`|`. Home and Library components owned their headings. `Seo.astro` already
handled OG metadata and Twitter title/description/card, but lacked
`twitter:image`; it emitted content language codes as `og:locale`. Publication,
robots, canonical, sitemap, hreflang and LanguageSwitcher membership were
already governed by the Phase 3A/3B manifest and remain so.

The shared pages/blog Zod schema now accepts optional `displayH1`. It must be a
string containing non-whitespace text. Omission preserves existing behavior;
empty, whitespace-only and non-string explicit values fail validation. Both
Front Matter CMS content types expose the field. No author or social-specific
schema fields were added.

- BlogPostLayout selects `displayH1 ?? title` for its existing H1 only.
- PageLayout selects `displayH1 ?? bannerTitle` only at PageHero. All existing
  heading maps and fallbacks remain intact. Form and contact heading props
  continue to receive the original banner title, avoiding collateral changes.
- HomePage passes the optional override to HomeHero. Without one, existing
  source heading HTML, including line breaks, is preserved. An override is
  escaped plain text and replaces the single H1, never adds another one.
- BlogIndexLayout reads the current manifest entry for an explicit H1 override.
  Its existing title prop continues to control the document/social titles and
  default H1. Existing locale-specific Library template titles are unchanged.

Cards, navigation labels, related-post text, article bodies, styles and client
code are unchanged. No content routes, policy fields or translation pairs were
added. The unpublished Arabic homepage remains unpublished.

### Newly Verified Heading Evidence

Two batches of read-only production requests on 2026-10-08 respected robots
and spaced page requests by at least one second. Nine GETs total comprised two
robots checks and seven article checks across four unique article URLs. All
returned HTTP 200; no redirects were followed, forms submitted, APIs called or
production assets downloaded. Three article checks were repeated with the
document-title selector restricted to `head title`, excluding inline SVG title
elements from the initial extraction. Only the corrected document titles below
are used as heading evidence. No production HTML captures are committed.

| Production URL | Observation UTC | Document title | Visible H1 | Local action |
| --- | --- | --- | --- | --- |
| [English self-advocacy](https://www.azinstitute4autism.com/library/autism-self-advocacy-skills-aba) | 2026-10-08 21:11:28 | Unlock Independence for Children with Autism Through ABA Self‑Advocacy | Guide to Teaching Self‑Advocacy in ABA | Add this `displayH1` only to `src/content/blog/en/autism-self-advocacy-skills-aba.md` |
| [Spanish self-advocacy](https://www.azinstitute4autism.com/es/library/autism-self-advocacy-skills-aba) | 2026-10-08 21:11:29 | Desbloqueando la independencia para niños con autismo a través de la autodefensa ABA | Guía para enseñar la autodefensa en el ABA | Add this `displayH1` only to `src/content/blog/es/autism-self-advocacy-skills-aba.md` |
| [Arabic self-advocacy](https://www.azinstitute4autism.com/ar/library/autism-self-advocacy-skills-aba) | 2026-10-08 21:11:30 | دليل تعليم مهارات الدفاع عن النفس في ABA | دليل تعليم مهارات الدفاع عن النفس في ABA | No change; existing title-to-H1 fallback is correct |

The English nonbreaking hyphen and Spanish heading are copied from the current
source, not newly translated. These are the only two content-record edits;
each adds one frontmatter line. Titles, descriptions, images, bodies and
translation declarations are untouched. This is not a linguistic review of
the translations or a broader certification of content currency.

A separate targeted check of
[`/library/aba-school-readiness-arizona`](https://www.azinstitute4autism.com/library/aba-school-readiness-arizona)
at **2026-10-08 21:07:17 UTC** found current OG title
“ABA School Readiness in Arizona: A Parent Guide” and H1
“ABA School Readiness for Autistic Children in Arizona.” The existing local
title is “ABA School Readiness in Arizona | AIA Preparatory Academy (Ages 2–6).”
The current live description also discusses classroom participation and
individualized ABA. This is fresh evidence of broader drift requiring later
content reconciliation; that record is intentionally unchanged in Phase 4A.

### Social Metadata and Locale

`Seo.astro` uses a small `socialImageUrl` helper to resolve existing image paths
against the configured public origin in `src/data/site.json`. It never uses a
preview request host or an article's external canonical as the asset base.
Absolute HTTP(S) image URLs are supported; blank selections produce no image,
and invalid URLs, non-HTTP(S) schemes or embedded credentials fail validation.

Image pages now emit matching absolute `og:image` and `twitter:image` once
each, retaining `summary_large_image`. Non-image pages emit neither image nor
image-alt metadata and retain `summary`. Existing image selection is unchanged:
article featured images and regular-page featured images/banner fallbacks are
preserved. No homepage image is invented merely because a hero image exists.
The real artifact has **90 image cards and six summary cards** in each mode.
Every emitted local social image resolves to an existing public asset.

BaseLayout forwards existing `alt` only when the selected image is the entry's
`featuredImage`. Seo emits both `og:image:alt` and `twitter:image:alt` only for a
nonblank description with a selected image. An unrelated banner fallback never
inherits that description. The current local pages/blog corpus has no
nonempty featured-image `alt` values, so no real image-alt tags are fabricated;
isolated fixtures prove the supported path in all three languages. Production
self-advocacy pages do have social alt descriptions. Importing those into local
content and reviewing visible image descriptions remain later content work.

`og:title`, `og:description`, `og:url`, `og:type`, Twitter title/description and
card selection otherwise retain their behavior. Articles remain `article` and
other pages `website`. No structured-data types were added.

The [Open Graph protocol](https://ogp.me/) defines the optional `og:locale` as
`language_TERRITORY`. Existing generic `en`, `es` and `ar` values were removed;
neither the frontmatter nor inspected production articles establish regional
assignments, and those production articles emit no `og:locale`. Omission avoids
inventing regions. Open Graph documents `en_US` as its default when omitted,
so intentional Spanish/Arabic social-region targeting remains an unresolved
human localization choice, not a claim that omission supplies it. HTML `lang`,
RTL/LTR direction, hreflang and `x-default` are unchanged.

### Validation

`npm run test:seo`: **12 passed**. Tests use the actual content schema, layouts,
SEO component and Astro configuration in temporary copies, covering:

- Original article H1 fallback and explicit override in en/es/ar; independent
  document/social/card titles; both production-supported records.
- Existing service, regular-page and derived headings; single H1; homepage
  inline heading markup and explicit home/Library overrides.
- Absolute root/path-relative and external image URLs, invalid/blank URLs,
  preview-host isolation, paired image metadata, existing alt and absent alt,
  no-image fallback, article type and language/direction.
- Shared publication-policy results on every generated fixture route in
  staging/indexing: canonical, robots, sitemap, hreflang and switcher targets,
  including all noindex/external-canonical exceptions and absent `/ar`.
- Actual failed content builds for `displayH1` values `""`, spaces, newline/tab
  and a number. Markdown records and an MDX service record exercise rendering.

The first test run exposed a fixture path using `.md` for the existing MDX ABA
service record; that test-only path was corrected before continuing. There
were no remaining test failures. `npm run test:publication`: **33 passed**,
unchanged from Phase 3B.

Completed commands, from `www/`:

```sh
npm run test:seo
npm run test:publication
npm run build
PUBLIC_ALLOW_INDEXING=true npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
npm run audit:links
npm run audit:blog
npm run audit:images
npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
git diff --check
```

Each normal/indexing build checked 73 files with **zero errors, warnings or
hints**, and generated **96 HTML routes**: 67 English, 21 Spanish, eight Arabic.
Link audit: **zero broken internal links**. Blog audit: **zero failures**.
Image audit: **zero missing mapped page images**; the generated social-image
check additionally verified all selected local image files. These are local
artifact checks, not verification of every external URL or social-platform
preview/cache behavior.

| Artifact | Checked UTC, 2026-10-08 | HTML routes | Sitemap URLs | Robots | Hreflang |
| --- | --- | ---: | ---: | --- | --- |
| Indexing-enabled local build | 21:16:30 | 96 | 92 | 92 `index,follow`; four `noindex,follow` | 174 links on 53 routes, same 24 explicit sets |
| Final normal staging build | 21:17:16 | 96 | 0 | All 96 `noindex,nofollow` | Zero links |

All 96 canonicals match the pre-change artifact and the shared policy. The
four exceptions remain `/schedule-consultation`, `/employee-portal`,
`/library/community-highlight-meet-rula-diab` and
`/library/new-aia-scottsdale-office`. Neither they nor unpublished `/ar` enter
the alternate graph or production-policy sitemap. Social URLs do not affect
route eligibility or canonical selection.

An all-route semantic comparison against the pre-change staging artifact found
exactly the two supported H1 changes. Remaining body markup, navigation and
scripts match after normalizing whitespace and existing random contact IDs;
H1 text is compared separately. Remaining head markup also matches after
excluding the intended social-tag differences and mode-dependent robots and
hreflang. No title, description, existing OG image, URL, direction, client
behavior or JSON-LD changes were introduced.

The final durable reconciliation is
[`route-reconciliation-2026-10-08-offline.json`](route-reconciliation-2026-10-08-offline.json).
It refreshes the application fingerprint and final staging output using the
same 70 saved production requests and their original observation dates. No
route-discovery network requests were made. The new targeted heading checks
above supplement that historical discovery evidence rather than overwrite it.
Counts remain 129 discovered routes, 102 historical production sitemap routes,
96 local routes, 92 overlapping routes and 10 production-sitemap routes absent
locally. The historical 102 is distinct from the local policy's 92. The
intermediate indexing reconciliation was moved to temporary storage before
the final staging reconciliation; prior committed evidence files are intact.

### Read-only Structural Review and Scope

The complete Phase 4A change set contains 17 files:

| Classification | Files relative to `www/` | Reason |
| --- | --- | --- |
| Schema/editor configuration | `src/content.config.ts`, `frontmatter.json` | Optional validated heading field on pages/blog and CMS exposure |
| Presentation | `src/layouts/BlogPostLayout.astro`, `src/layouts/PageLayout.astro`, `src/layouts/BlogIndexLayout.astro`, `src/components/pages/HomePage.astro`, `src/components/home/HomeHero.astro` | Explicit single-H1 override, preserving each template's default |
| SEO | `src/components/Seo.astro`, `src/layouts/BaseLayout.astro`, **new** `src/utils/seo.ts` | Paired absolute social images, conditional existing alt, removal of unsupported generic OG locales |
| Tests | **new** `tools/seo.test.mjs`, `package.json` | Actual Astro/schema regression coverage and `test:seo`; no dependency changes |
| Content metadata | `src/content/blog/en/autism-self-advocacy-skills-aba.md`, `src/content/blog/es/autism-self-advocacy-skills-aba.md` | One source-supported frontmatter line per record |
| Documentation/evidence | `README.md`, this report, **new** `reports/route-reconciliation-2026-10-08-offline.json` | Editing rules, dated evidence/results, current offline fingerprint |

No changes to publication-policy code, route dispatch, LanguageSwitcher,
Header, styling, dependency versions/lockfile, environment or deployment
configuration are required. Existing audit-generated link/blog/image reports
are unchanged. Employee-portal implementation, the separate Ads project and
all unrelated files remain untouched. The pre-existing untracked
`merge-plan.md` retains SHA-256
`015db80cdbaf7d68799265d2070db760155c342b432ce6795f17ebc2218c641c`.

No blocking Phase 3A/3B defect was found. Deferred work includes broader title,
content and image-alt reconciliation, regional social locale decisions,
article/FAQ schema, and human language review where substantive translations
change. No Phase 4B work or commit is authorized by this implementation review.

## Phase 4B — Article and FAQ Structured Data (2026-10-08)

Implemented on `faithful-astro-migration`, starting from
`4ed01d7d288b249bd028a492e534d3d91adad6f7`. The working tree initially contained
only the unrelated untracked `merge-plan.md`. The current history has the
Phase 4A change set after two Ads documentation commits; this phase does not
alter or incorporate that handoff work. Earlier milestone sections remain
historical evidence. Nothing in this phase is staged or committed.

### Existing Implementation and New Production Evidence

Inspection covered BlogPostLayout, BaseLayout, FAQAccordion, FaqPage, all FAQ
component consumers, content schemas, the three author records, publication
helpers, the existing SEO tests, and the plan's Phase 4B acceptance criteria.
The initial generated artifact had 96 routes, one existing MedicalOrganization
block per route, no BlogPosting blocks, and 20 FAQPage blocks: 125 questions in
19 English/Spanish MDX articles and two insurance questions. Those 127 schema
answers matched their rendered answers after whitespace normalization.

`/faqs` was different: its rendered Markdown/MDX source contained 70 unique
direct H3 questions, but FAQAccordion received no `items`. Browser code grouped
the H3s and following siblings into accordions at runtime, so the server emitted
no FAQPage. This was a missing schema path, not a reason to replace the working
accordion, its interaction code, or its content.

Nine read-only production GETs on **2026-10-08 21:57:43–21:57:53 UTC** comprised
robots first and the eight targeted pages below. Robots allowed these paths;
all responses were HTTP 200, requests were at least one second apart, and no
redirects were followed. No forms, APIs, production writes or broad crawl were
used. Production HTML stayed in temporary storage.

| Production page | Observation UTC | Relevant findings |
| --- | --- | --- |
| [English self-advocacy](https://www.azinstitute4autism.com/library/autism-self-advocacy-skills-aba) | 21:57:44 | BlogPosting headline uses the SEO title, not display H1; Rula Diab author; published `2024-12-02T07:00:00.000Z`, modified `2026-10-03T20:47:43.625Z` |
| [Spanish self-advocacy](https://www.azinstitute4autism.com/es/library/autism-self-advocacy-skills-aba) | 21:57:45 | BlogPosting, localized Rula Diab byline; published `2026-04-09T03:36:24.000Z`, modified `2026-04-09T03:36:32.552Z`; current visible date is April 9, 2026 |
| [Arabic self-advocacy](https://www.azinstitute4autism.com/ar/library/autism-self-advocacy-skills-aba) | 21:57:47 | BlogPosting, localized Rula Diab byline; published `2024-12-02T07:00:00.000Z`, modified `2025-09-09T01:34:16.718Z` |
| [Standalone FAQs](https://www.azinstitute4autism.com/faqs) | 21:57:48 | One FAQPage with 70 questions |
| [Insurance](https://www.azinstitute4autism.com/insurance) | 21:57:49 | One FAQPage with two questions; local question/answer text matches the production schema after normalization |
| [School readiness](https://www.azinstitute4autism.com/library/aba-school-readiness-arizona) | 21:57:51 | BlogPosting plus FAQPage with nine questions; local article has seven older questions |
| [Community interview](https://www.azinstitute4autism.com/library/community-highlight-meet-rula-diab) | 21:57:52 | `noindex`, external Voyage Phoenix canonical; HubSpot still supplies BlogPosting with AIA-logo publisher and Rula Diab author |
| [Office article](https://www.azinstitute4autism.com/library/new-aia-scottsdale-office) | 21:57:53 | `noindex`, external Scottsdale Airpark canonical; same generic HubSpot authorship/publisher pattern |

Current source checks support the shared metadata structure and selected
records; they do not certify the currency of all 65 local articles. Identified
date/copy drift is explicitly deferred below, not silently reconciled.

### Article Architecture and Publication Exceptions

`src/utils/structured-data.ts` holds the schema builders and an HTML-safe JSON
serializer. BlogPostLayout consumes the existing manifest publication record
and authors collection and emits one BlogPosting through BaseLayout's existing
head slot. BaseLayout's MedicalOrganization block is unchanged.

BlogPosting includes canonical `@id` (`#article`), `url`, `mainEntityOfPage`, the
existing SEO `title` as `headline`, language, declared publication day, and
optional nonblank description, featured image and explicit modification day.
Image resolution reuses Phase 4A's public-origin helper. Display H1 does not
replace the SEO title. No body text, keywords, publisher, credentials, original
publication claim or missing image is fabricated.

Author resolution requires exactly one record with the article's declared
author slug and language. Its `name` supplies a Person name. Missing/ambiguous
references fail, rather than silently substituting Rula Diab or an English
record. Existing authors all identify Rula Diab. Author URLs are omitted because
the repository does not yet generate author archives. No author records or
frontmatter/schema fields changed.

Dates use the precision actually supplied by current records: calendar days.
`updatedDate` is emitted only when explicitly present; all current records lack
it. Neither build time, Git dates nor recently observed HubSpot update times
are substituted. Modification before publication fails validation.

Both article and FAQ builders reuse the manifest's **record-level
`sitemapEligible`** gate. Drafts, noindex and external-canonical records emit
neither new type. The two syndicated articles above therefore receive no
invented original author/publisher claim, even though production HubSpot emits
generic article schema there. Their external canonicals, noindex and global
site organization block remain intact. `/schedule-consultation` and
`/employee-portal` likewise emit neither type; `/ar` remains unpublished.

The record gate is independent of `PUBLIC_ALLOW_INDEXING`. Eligible records
emit schema in both build modes for local testing, while staging retains its
global noindex protection, empty sitemap and absent hreflang. Schema does not
make a record indexable or create translation relationships.

### FAQ Verification and Corrections

FAQAccordion remains the single FAQ schema producer. Its existing supplied-item
markup, CSS, animation, search and runtime grouping code are preserved.

- Schema answers derive from the HTML actually rendered by the component,
  using `answerHtml` when present or the same `answer` fallback. This prevents
  a stale duplicate plain-text field from contradicting the visible answer.
  Entity decoding and block separators preserve readable text; nonvisible
  script/style/template and explicitly hidden nodes are excluded.
- FaqPage renders its existing slot once, displays that same HTML, and passes
  it to FAQAccordion as `sourceHtml`. The builder groups direct H3 questions
  and following siblings just as the unchanged browser enhancement does.
  Nested H4/H5 headings and lists stay inside their answer. All 70 questions
  are now described in server JSON-LD, including with JavaScript disabled.
- A runtime-only selector without server source emits no invented FAQ schema.
  Supplied items cannot be mixed with a runtime source. Empty groups emit none.
- Schema IDs use the manifest canonical. Any retained component canonical prop
  must agree with it; conflicting values fail. Existing canonical props and
  article content are unchanged.
- Empty questions/answers and duplicate questions fail. A request-scoped
  `Astro.locals` claim rejects a second FAQ schema-producing accordion on the
  same page. Separate rendered pages do not share this state. Existing
  organization, article and FAQ blocks coexist with distinct entity IDs.
- JSON-LD escapes `<` so script-like text cannot terminate its HTML script
  element. This is exercised in both helper and rendered-article fixtures.

Result: **21 FAQPage blocks, 197 questions** (125 article, two insurance, 70
standalone), and **63 BlogPosting blocks** (65 articles minus two syndicated
exceptions). No unrelated schema types or new organization entities were added.

### Date-Display Defect Found During Validation

The build host uses `America/Phoenix`. Existing article/card date formatting
converted date-only frontmatter through that timezone: declared `2024-12-02`
appeared as December 1 despite the machine-readable time retaining December 2.
That would make the new schema disagree with the displayed calendar day.

The minimal correction adds `timeZone: 'UTC'` to the two existing article/card
formatters. It preserves declared dates, formatting options and locale, and
does not update any article to match a newer production publication date. On
this host it corrects **130 labels**: 65 article dates and 65 Library-card dates.
SEO tests explicitly build under `America/Phoenix` and compare both visible
formats with each record's declared date in all three languages. This is a
rendering-consistency fix within Phase 4B, distinct from Phase 6C date research.

### Phase 6C Drift and Remaining Uncertainty

- Spanish self-advocacy still declares `2024-12-02` locally, while the current
  production article declares/displays April 9, 2026. Its local publication
  record was not rewritten. Reconcile translation publication history in
  Phase 6C; do not mistake the timezone repair for this separate correction.
- English school-readiness has seven older local FAQ items around the former
  academy/program description. Production has nine questions about current
  school-readiness skills and ABA's relationship to school. Its shared academy
  question also has a different answer. Title/content/FAQ reconciliation stays
  in Phase 6C; no answers or Library bodies changed here.
- Production article modification timestamps and localized bylines are richer
  than current local records. They were observed but not imported into otherwise
  unreconciled records. No claim of full current-content parity is made.
- The 70 standalone questions are derived from actual rendered local source.
  Text extraction shows punctuation/HTML-spacing differences from production;
  this milestone is not a sentence-by-sentence clinical/content review.

Schema.org still defines [BlogPosting](https://schema.org/BlogPosting) and
[FAQPage](https://schema.org/FAQPage). Google's
[article documentation](https://developers.google.com/search/docs/appearance/structured-data/article)
supports applicable, source-backed properties without requiring fabricated
optional fields. Its [current changelog](https://developers.google.com/search/updates)
records removal of FAQ rich results beginning May 7, 2026 and removal of that
feature's documentation in June. This implementation preserves semantic FAQ
markup and source fidelity; it makes no promise of Google FAQ rich results.

The browser tool returned `Transport closed`, so interactive desktop/mobile
checks could not run. The accordion's client code and CSS are unchanged, and
all-route rendered comparisons verify content/markup, scripts and style content.
Browser interaction/visual confirmation and external rich-result tooling remain
manual checks; no remote validator submission or deployment was performed.

### Validation and Artifact Invariants

Final `npm run test:seo`: **22 passed**, comprising the 12 Phase 4A tests plus
10 structured-data/date tests. Coverage includes en/es/ar articles, absent
optional metadata, declared/modified dates, exact author resolution, noindex,
external canonical and draft exclusions, every rendered FAQ answer, the
standalone runtime source, empty/unknown sources, multiple JSON-LD blocks,
duplicate/conflicting schema failures, safe serialization and timezone behavior.
`npm run test:publication`: **33 passed** with no changes to that suite.

An initial standalone-FAQ assertion incorrectly compared sibling text without
block-boundary whitespace. The assertion was corrected and extraction now
explicitly preserves block separators. After the date-display defect was found
and fixed, the full tests/build/audit sequence was rerun successfully:

```sh
cd www
npm run test:seo
npm run test:publication
npm run build
PUBLIC_ALLOW_INDEXING=true npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
npm run audit:links
npm run audit:blog
npm run audit:images
npm run build
npm run audit:routes -- --offline
npm run audit:routes -- --check
git diff --check
```

Both normal builds and the indexing-enabled build checked 74 files with zero
errors, warnings or hints. Link, blog and mapped-image audits each reported
zero failures. All selected article schema images resolve to local public assets.

| Final artifact | Verified UTC, 2026-10-08 | HTML routes | Sitemap URLs | Robots | Hreflang |
| --- | --- | ---: | ---: | --- | --- |
| Indexing-enabled local | 22:21:40 | 96 | 92 | 92 `index,follow`; four `noindex,follow` | 174 links on 53 routes, unchanged 24 sets |
| Normal staging, restored | 22:22:24 | 96 | 0 | All 96 `noindex,nofollow` | Zero links |

The route set remains 67 English, 21 Spanish and eight Arabic. All 96 canonicals,
non-schema metadata, H1s, HTML language/direction and translation choices match
the starting artifact/policy. Existing MedicalOrganization objects match on
all 96 routes. Body markup and scripts match after accounting for the explicit
date-label fix, added JSON-LD, whitespace and pre-existing random contact IDs.
All inline CSS declarations and stylesheet URLs are unchanged. Astro moved the
unchanged FAQ stylesheet link relative to an unrelated inline insurance-style
block after making FAQ rendering asynchronous; no selectors/declarations changed.

The final durable evidence is
[`route-reconciliation-2026-10-08-offline-22-22-25-260Z.json`](route-reconciliation-2026-10-08-offline-22-22-25-260Z.json).
It reuses all 70 historical discovery requests with their observation times.
Route rows/totals remain unchanged: 129 discovered routes, historical production
sitemap 102, local generated 96, overlap 92, and ten historical sitemap routes
missing locally. Intermediate indexing evidence stayed in temporary storage;
the prior committed audit evidence remains intact. No fresh crawl was made to
refresh fingerprints; the targeted schema observations above are separately dated.

### Read-only Structural Review

The complete Phase 4B change set has nine files, all relative to `www/`:

| Classification | File | Purpose |
| --- | --- | --- |
| Structured-data core, new | `src/utils/structured-data.ts` | Publication-gated builders, author validation, visible FAQ extraction, safe JSON serialization |
| Article presentation/schema | `src/layouts/BlogPostLayout.astro` | Typed article entry, head-slot BlogPosting, calendar-day formatting |
| Date presentation | `src/components/BlogCard.astro` | Same calendar-day correction for Library cards |
| Existing FAQ producer | `src/components/FAQAccordion.astro` | Shared builder, manifest canonical validation, rendered-source support, duplicate guard; client script/styles unchanged |
| Runtime FAQ source bridge | `src/components/pages/FaqPage.astro` | Render slot once for both visible source and server schema |
| Regression coverage | `tools/seo.test.mjs` | Extend existing actual-Astro tests; no parallel publication policy |
| Editing documentation | `README.md` | Schema/date/FAQ rules and known reconciliation boundaries |
| Milestone evidence | `reports/migration-summary.md` | This dated implementation and review record |
| Offline evidence, new | `reports/route-reconciliation-2026-10-08-offline-22-22-25-260Z.json` | Final source fingerprint with preserved production discovery |

No content/frontmatter records, content schemas, dependencies, publication or
translation policy, routing, deployment files, employee-portal work or Ads
handoff files were modified. `merge-plan.md` remains untracked and unchanged,
SHA-256 `015db80cdbaf7d68799265d2070db760155c342b432ce6795f17ebc2218c641c`.
There is no new blocking policy question; Phase 6C must resolve the documented
content/date drift. Stop for implementation/diff acceptance before any checkpoint
commit. Phase 5A has not started.

## Phase 5A — Services, Referrals, and Consultation (2026-10-09 UTC)

Implemented from Phase 4B checkpoint
`4079156776dd6b5a1b4e42b02dd9f680944484dd` on
`faithful-astro-migration`. This section records the implementation and read-only
review; nothing is staged or committed. The shell's actual UTC retrieval date
was October 9. AIA's governing editorial confirmation is dated **October 8,
2026**, as supplied by the user. Earlier reports and production observations
remain historical evidence, not competing authority for the confirmed correction.

### Scope, Authority, and Source Freshness

Inspected all seven English primary pages and all six existing Spanish
counterparts: ABA therapy, autism evaluations, Learner Social Club, services,
client consultation, intake, and English referrals. No Spanish referrals record
exists; none was created. Also inspected English schedule consultation and
insurance; FormShell, ServiceInquiryDrawer, ServicesPage, PageLayout, Button,
shared service data and relevant navigation/CTA code. Publication, translation,
heading, social and structured-data infrastructure was inspected and retained.

AIA explicitly confirmed general ABA eligibility as **18 months through 8
years / 18 meses a 8 años** and superseded broader children/teens descriptions.
This decision governs the in-scope edits even when cached production content
conflicts. It is resolved, not a human-decision blocker. It does not establish
ages for the Academy, Social Club or diagnostic evaluations.

Durable new source evidence:
[`phase-5a-source-evidence-2026-10-09.json`](phase-5a-source-evidence-2026-10-09.json).
It records exact URLs, request times, HTTP status, request/response cache
indicators, body hashes, titles, descriptions, H1s, canonicals, selected source
excerpts and CTA destinations. It contains no complete HTML captures or cookies.

There were **23 successful read-only GETs**: AIA robots, 15 targeted AIA pages,
four cache rechecks, Jotform robots and two bare Jotform destination status
checks. Page requests were sequential with at least 1.1 seconds between them;
redirects were not followed. AIA robots allowed the selected paths. Jotform
robots permits the bare numeric paths; its disallowed tracking-query form
variant was not fetched. No scripts, forms, APIs, authenticated operations,
tracking endpoints or production writes were executed. A web-tool robots
attempt failed before the successful direct retrieval. The independent browser
attempt failed with `Transport closed`; no browser response was obtained.

Every AIA page request sent `Cache-Control: no-cache, max-age=0` and
`Pragma: no-cache`. Responses still carried Cloudflare/HubSpot cache indicators:
`Cache-Control: s-maxage=36000, max-age=5` and
`X-HS-Cache-Control: s-maxage=36000, max-age=0`. `Age`, page `ETag` and
`CF-Cache-Status` were absent; `X-HS-CF-Cache-Status`, `Last-Modified`,
`X-HS-Prerendered`, response Date and CF-Ray are saved where available.
Prerender dates matched Last-Modified. **HTTP 200, MISS and REVALIDATED do not
prove that the underlying prerendered content contains the newest edits.**

Initial source observations (all times UTC, October 9, 2026):

| Source URL | Request time | HTTP | X-HS-CF-Cache-Status | Last-Modified (GMT) |
| --- | --- | ---: | --- | --- |
| [/aba-therapy](https://www.azinstitute4autism.com/aba-therapy) | 01:26:21 | 200 | HIT | Sat, 03 Oct 2026 23:27:01 GMT |
| [/autism-evaluations](https://www.azinstitute4autism.com/autism-evaluations) | 01:26:22 | 200 | HIT | Sat, 03 Oct 2026 23:13:18 GMT |
| [/learner-social-club](https://www.azinstitute4autism.com/learner-social-club) | 01:26:23 | 200 | HIT | Sat, 03 Oct 2026 23:13:14 GMT |
| [/services](https://www.azinstitute4autism.com/services) | 01:26:25 | 200 | HIT | Sat, 03 Oct 2026 23:13:23 GMT |
| [/referrals](https://www.azinstitute4autism.com/referrals) | 01:26:26 | 200 | HIT | Sat, 03 Oct 2026 23:45:35 GMT |
| [/client-consultation](https://www.azinstitute4autism.com/client-consultation) | 01:26:28 | 200 | HIT | Sun, 04 Oct 2026 00:00:00 GMT |
| [/aba-therapy-intake-process](https://www.azinstitute4autism.com/aba-therapy-intake-process) | 01:26:29 | 200 | HIT | Sat, 03 Oct 2026 23:13:15 GMT |
| [/es/aba-therapy](https://www.azinstitute4autism.com/es/aba-therapy) | 01:26:30 | 200 | HIT | Sat, 03 Oct 2026 23:13:14 GMT |
| [/es/autism-evaluations](https://www.azinstitute4autism.com/es/autism-evaluations) | 01:26:32 | 200 | MISS | Sat, 03 Oct 2026 23:13:12 GMT |
| [/es/learner-social-club](https://www.azinstitute4autism.com/es/learner-social-club) | 01:26:33 | 200 | MISS | Sat, 03 Oct 2026 23:13:22 GMT |
| [/es/services](https://www.azinstitute4autism.com/es/services) | 01:26:35 | 200 | REVALIDATED | Sat, 03 Oct 2026 23:13:22 GMT |
| [/es/client-consultation](https://www.azinstitute4autism.com/es/client-consultation) | 01:26:36 | 200 | REVALIDATED | Sun, 04 Oct 2026 00:00:44 GMT |
| [/es/aba-therapy-intake-process](https://www.azinstitute4autism.com/es/aba-therapy-intake-process) | 01:26:38 | 200 | MISS | Sat, 03 Oct 2026 23:13:18 GMT |
| [/schedule-consultation](https://www.azinstitute4autism.com/schedule-consultation) | 01:26:39 | 200 | REVALIDATED | Sat, 03 Oct 2026 23:13:23 GMT |
| [/insurance](https://www.azinstitute4autism.com/insurance) | 01:26:41 | 200 | HIT | Sat, 03 Oct 2026 23:13:18 GMT |

Rechecks at 01:28:06–01:28:10 used the same headers and the permitted unique
query `?migration_review=20261009T0130` on `/aba-therapy`, `/es/aba-therapy`,
`/client-consultation` and `/schedule-consultation`. All four returned HIT with
identical extracted text and metadata to their initial responses. This does
not guarantee an origin refresh. In particular, schedule consultation still
returned “children and teens,” contradicting the newer AIA instruction; it is
classified as potentially stale. Other responses corroborate the approved ABA
ages and removal of teens from both client-consultation pages. Origin freshness
remains inconclusive overall. No production change is inferred from response
status alone.

### Page-by-page Reconciliation

In this table, **P** means a newly retrieved production response with origin
freshness inconclusive (see the exact request time above). **E** means AIA's
confirmed October 8 editorial instruction. P/E records do not assert cache
bypass success. CTA resolutions are detailed separately below.

| Route / language | Claim or element | Previous Astro state | Retrieved production state | Editorial requirement | Source time / freshness | Resolution and notes |
| --- | --- | --- | --- | --- | --- | --- |
| `/aba-therapy` en | Individual 1:1 age | “children aged 2 to 8 years” | “children aged 18 months to 8 years” | “18 months through 8 years” | P 01:26:21 + E; origin inconclusive, decision confirmed | Changed to “children aged 18 months through 8 years” |
| `/aba-therapy` en | Section heading | “Behavioral Services We Offer” | “Services We Offer” | — | P 01:26:21; origin inconclusive | Changed label; explicitly retain `behavioral-services-we-offer` ID to preserve existing fragment URLs |
| `/aba-therapy` en | Group program | 1:2 Academic Readiness, ages 2–6, AIA Preparatory Academy© | Same distinct program, ratios, range and branding | General age rule does not override this program | P 01:26:21 | Retained paragraphs, electives and 2–6 range; no clinical claims added |
| `/es/aba-therapy` es | General/1:1 eligibility | “personas de 2 a 17 años de edad, que han sido diagnosticadas”; “niños de 2 a 8 años” | General introduction says children; 1:1 says “niños de 18 meses a 8 años” | “18 meses a 8 años” | P 01:26:30 + E | Changed introduction to “niños de 18 meses a 8 años de edad, que han sido diagnosticados” and 1:1 to approved age phrase; preserves paragraph structure |
| `/es/aba-therapy` es | Academy branding | Link label “Academia Preparatoria AIA©” | Proper name “AIA Preparatory Academy©” | — | P 01:26:30; origin inconclusive | Use source-supported proper name; retain existing `https://aiaprep.org/` destination and 2–6 group-program range |
| `/autism-evaluations` en | Meta description | Ends at “support your child” | Ends at “support your child's unique development.” | — | P 01:26:22; origin inconclusive | Restore the complete source description; title and H1 unchanged |
| `/autism-evaluations` en; `/es/autism-evaluations` es | Diagnostic services | Separate ASD, ADHD and combined evaluation; ADOS discussion | Same offerings and clinical descriptions | Do not inherit ABA ages | P 01:26:22 / 01:26:32 | Body retained. “Children of all ages” describes the tool; it is not verified clinic eligibility. No clinic age limits, staffing or prerequisites invented |
| `/learner-social-club` en; `/es/learner-social-club` es | Age, schedule, payment and enrollment | Ages 8–17; 4–6 PM Mon/Wed or Wed; private pay; potty training; 30-day pause notice; one-day trial by email | Matches retrieved page content and metadata in each language | Separate program | P 01:26:23 / 01:26:33 | All facts retained. Only enrollment CTA destination changes; availability still requires confirmation before activation |
| `/services` en; `/es/services` es | Services, geography/options and clinical copy | In-center/school/home, early intervention, academic readiness, supervision, parent consultation; “1 in 36 … each year” | Matches retrieved content, including the prevalence wording | Do not manufacture service availability | P 01:26:25 / 01:26:35 | Retained. English body is owned by ServicesPage, not its MDX slot. No need to edit either English duplicate. Clinical statistic held for review rather than silently replacing it |
| `/es/services` es | Home breadcrumb | `/` | `/es/?hsLang=es` | Keep published language routing | P 01:26:35 | Changed to `/es`; existing Spanish service/consultation links remain localized, rather than copying production English paths with `hsLang=es` |
| `/referrals` en | Referral audience | “educational and therapeutic programs for children ages 2-17” | Same sentence with “18 months to 8 years” | Confirmed range is general ABA, not every program | P 01:26:26 + E | “ABA therapy programs for children ages 18 months through 8 years”; explicitly limits the claim to ABA. Existing provider/insurer collaboration and reporting descriptions retained |
| `/client-consultation` en | Audience/introduction | “Here at”; two “children and teens” references | “At”; both references say “children” | Approved ABA age and audience | P 01:26:28 + E | Source intro “At”; remove teens; ABA services sentence specifies “children aged 18 months through 8 years.” No-wait-list and existing Phoenix-area geography retained pending operational confirmation |
| `/es/client-consultation` es | Audience | Two “niños y adolescentes” references | Both say “niños” | Approved Spanish ABA age/audience | P 01:26:36 + E | Remove “y adolescentes”; ABA sentence specifies “niños de 18 meses a 8 años.” Remaining Spanish copy unchanged |
| `/aba-therapy-intake-process` en; `/es/aba-therapy-intake-process` es | Intake steps and CTA | Six steps; verification of insurance, copays and required authorizations; final label is plain text | Same six-step process with final consultation link | No new universal prerequisites | P 01:26:29 / 01:26:38 | Steps and insurance wording retained; restore styled linked CTA using existing local consultation pages. No intake packet/submission functionality added |
| `/schedule-consultation` en | Audience | Two “children and teens” references | Same superseded wording despite recheck | E overrides stale response | P 01:26:39 / recheck 01:28:10, potentially stale + E | Remove teens and explicitly state ABA age range in ABA services sentence. Retain noindex, existing geography/contact details and disabled form; do not treat response as fresh editorial authority |
| `/insurance` en | Plans/payment and FAQ | Logo list and existing ECHO/ESA FAQ claims | Same visible carrier logos/FAQ text | No plan-to-plan generalization | P 01:26:41; origin inconclusive | Unchanged; independently compared this page rather than assuming service/form fields establish every plan's coverage. Participation/coverage details remain unverified |

No article, FAQ answer, homepage, team or portal content was changed. No new
Spanish records, translations or translation keys were introduced. All page
titles and visible H1s remain unchanged. The evaluation description is the only
frontmatter value changed. Changes to metadata values use the existing SEO
pipeline; no schema-generation rule or organization data changed.

### CTA Reconciliation and Disabled Forms

Bare production Jotform destinations `231638510219149` and `250727122848156`
returned HTTP 200 at 01:30:34 and 01:30:35 respectively. These were GET status
checks only, with response bodies/scripts unused; availability does not approve
collection or establish form field/backend equivalence. The referral query
variant was not requested because of Jotform robots policy.

| Route(s) | Production label / destination | Previous Astro label / destination | Implemented migration destination / verification | Integration dependency |
| --- | --- | --- | --- | --- |
| `/aba-therapy`, `/autism-evaluations` | “Make an Appointment” → `https://form.jotform.com/231638510219149` | Same label and external URL | Same label → `/client-consultation`; generated destination exists, live consultation 200 | Existing disabled consultation shell; this is an inquiry path, not a booked appointment |
| `/es/aba-therapy` | “Hacer una cita” → same Jotform | Same label as unlinked text | Same label → `/es/client-consultation`; generated/live 200 | Same disabled shell; human Spanish form review still needed |
| `/es/autism-evaluations` | “Programar una cita” → same Jotform | Same label as unlinked text | Same label → `/es/client-consultation`; generated/live 200 | Evaluation intake/backend not implemented |
| Both Learner Social Club routes | “Enroll Now” → same Jotform (English label also on Spanish production) | Same label/external URL | Same label → corresponding English/Spanish consultation page; both resolve | Does not enroll or reserve a place; backend and service-specific process remain gated |
| Both Social Club routes, trial inquiry | One-day trial via obfuscated `info@azinstitute4autism.com` | Same address/ContactObfuscation behavior | Retained; no email sent | Trial availability must be confirmed by AIA |
| `/services` | “Consult with a Client Advocate” / “Get a Free Consultation” → `/client-consultation?hsLang=en` | Same labels → `/client-consultation` | Retained normalized local destination; resolves | Disabled consultation shell |
| `/es/services` | “Consulte con un Defensor del Cliente” → `/es/client-consultation?hsLang=es`; “Programe una consulta gratuita” → English path with `hsLang=es` | Both labels → `/es/client-consultation` | Retain existing real Spanish destination; resolves | Do not guess language from a query parameter |
| `/referrals` | “Refer a Client” → `https://form.jotform.com/231638510219149?hsCtaAttrib=203953990377` | Unlinked “Refer a Client”; existing disabled FormShell below | Same label → `#form-title`; exactly one matching element in generated page | Safe static referral placeholder; clinical referral workflow/required fields still need approval |
| Both intake routes | “Take the First Step” → `/client-consultation?hsLang=en` or `?hsLang=es` | Same unlinked label | `/client-consultation` or `/es/client-consultation`; both resolve | Starts inquiry, does not submit the described learner packet |
| Both client-consultation routes | Embedded Jotform `250727122848156` | Disabled local FormShell | Unchanged disabled shell; no embed, submit or success state added | Production form/backend/privacy approval outstanding |
| `/schedule-consultation` and service drawers | Production HubSpot/runtime form behavior; not executed | Disabled forms and call fallback | Retained; submission buttons disabled, no action destination, existing preventDefault handlers | Field parity, availability, clinical routing and backend require separate scoped activation |

The intentional temporary deviation is routing conversion links to existing
local disabled inquiry pages instead of third-party lead collection. Existing
form text explicitly states “Online submission is not yet connected” and offers
the existing phone fallback. No functioning submission is implied by the target.
Buttons retain source labels (including untranslated labels already present in
Spanish production); no new marketing or translation wording was invented.

Five `.md` records became `.mdx` because they now genuinely embed the existing
Button component: English referrals and intake, Spanish ABA, evaluations and
intake. Their collection IDs/slugs/public routes do not change. No routing or
layout conditionals were added. Existing Button styling is reused. FormShell,
ServiceInquiryDrawer, client scripts, fields, privacy behavior and form handlers
are unchanged. Production promises about callbacks/availability are retained
as source copy, not enabled functionality; they need review before activation.

### Repository-wide Eligibility Language Audit

Searched the full application source (pages/blog/authors, frontmatter, schema,
components, shared data and navigation), plus CMS configuration, for hyphen/en
and em-dash ranges, “to”/“a”/“hasta,” ages/years/months, Spanish niños/adolescentes,
English children/teens and Arabic age/adolescent terms. A broad semantic pass
produced 204 candidate lines, including non-eligibility words in code. Context
review excludes biographies, testimonials, generic developmental examples,
clinical research ages and historical reports from current AIA eligibility
corrections. Prior reports and baseline JSON are preserved, not globally replaced.
Locations below refer to the Phase 4B starting checkpoint unless marked current.

| Classification | File / location (relative to `www/`) | Exact relevant wording | Meaning, resolution and assigned phase |
| --- | --- | --- | --- |
| Corrected within Phase 5A | `src/content/pages/en/aba-therapy.mdx:27` | “children aged 2 to 8 years” | ABA 1:1 → confirmed 18 months through 8 years |
| Corrected within Phase 5A | `src/content/pages/es/aba-therapy.md:12,24` (now `.mdx`) | “personas de 2 a 17 años de edad”; “niños de 2 a 8 años” | General ABA/1:1 → approved Spanish range and agreement |
| Corrected within Phase 5A | `src/content/pages/en/referrals.md:14` (now `.mdx`) | “programs for children ages 2-17” | Referral ABA eligibility → confirmed range, explicitly scoped to ABA |
| Corrected within Phase 5A | `src/content/pages/en/client-consultation.md:10,14` | “children and teens”; “ABA therapy for children and teens” | Remove superseded audience; explicit ABA range |
| Corrected within Phase 5A | `src/content/pages/es/client-consultation.md:10,14` | “niños y adolescentes”; “terapia ABA para niños y adolescentes” | Remove superseded audience; explicit approved Spanish range |
| Corrected within Phase 5A | `src/content/pages/en/schedule-consultation.mdx:33,37` | “children and teens diagnosed with autism” | Correct despite potentially stale retrieval; noindex retained |
| Legitimate program-specific requirement | Both ABA page records, group-program paragraph (en:37, es:34 before rename) | “children aged 2 to 6 years”; “niños de 2 a 6 años” | Academic Readiness 1:2, supported by separate live program sections; retained |
| Legitimate program-specific requirement | Both Social Club records, descriptions and en:43 / es:50 | “8–17 years”; “jóvenes de 8 a 17 años” | Club ages, separate from ABA; retained, with private-pay distinction |
| Legitimate program-specific requirement | `src/data/services.json:16` | “Social and emotional learning opportunities for children and teens.” | Sociological card links specifically to the 8–17 Club; not general ABA; retained |
| Potentially outdated, deferred to another phase | `src/content/pages/en/index.md:14,23` | “clinical care for children and teens who have an autism diagnosis”; “ABA Therapy Services for Children and Teens Diagnosed with Autism Spectrum Disorder” | General ABA/homepage claims; **Phase 5B**, untouched |
| Potentially outdated, deferred to another phase | `src/content/pages/es/index.md:14,23` | “atención clínica experta para niños y adolescentes con diagnóstico de autismo”; “Servicios de terapia ABA para niños y adolescentes diagnosticados con TEA” | Spanish homepage general ABA claims; **Phase 5B**, source/language review in **Phase 7** as needed; untouched |
| Potentially outdated, deferred to another phase | `src/content/blog/en/aba-therapy-summer-routine-tips.mdx:20,149` | “resources tailored for children ages 2–17”; “In-clinic ABA therapy for children ages 2–8” | General AIA/ABA claims; **Phase 6C**. Club-specific “In-clinic Social Learners Club for ages 8-17” at :151 is separate and retained |
| Potentially outdated, deferred to another phase | `src/content/blog/en/community-highlight-meet-rula-diab.md:31,33,39` | “clinical care for children and teens diagnosed with Autism Spectrum Disorders”; “Our range of services for children and teens includes:”; “ABA Therapy Services for Children & teens” | Syndicated interview/ABA claims; **Phase 6C**, including original-source/editorial treatment; external canonical and noindex preserved |
| Potentially outdated, deferred to another phase | `src/content/blog/en/aba-school-readiness-arizona.mdx:2,3,18,154,155` | “AIA Preparatory Academy (Ages 2–6)”; “We currently serve children ages 2 to 6 who are in their most critical years of early development.” | Academy-specific range is not automatically wrong, but article was rewritten live (Phase 4B evidence). **Phase 6C**, not globally changed to ABA range |
| Potentially outdated, deferred to another phase | `src/content/blog/es/aba-school-readiness-arizona.mdx:2,3,18,72,154,155` | “Academia Preparatoria AIA (2 a 6 años)”; “Actualmente atendemos a niños de 2 a 6 años que se encuentran en sus años más críticos de desarrollo temprano.” | Same program/context distinction; **Phase 6C**, human language review/**Phase 7** where needed |
| Potentially outdated, deferred to another phase | `src/data/blog-footers.json:352,551` and `src/content/pages/es/library.md:14` | “Ya se inscriben alumnos: Academia Preparatoria AIA para niños de 2 a 6 años” | Related/popular Library titles reflecting the older article; **Phase 6C**, untouched |
| Potentially outdated, deferred to another phase | `src/data/blog-footers.json:629,749,948,959,1237,1657,1671,1724,1761` | “Now Enrolling: AIA Preparatory Academy for Children Ages 2–6” | Library links/title drift, not proof of wrong program eligibility; **Phase 6C** |
| Unresolved and requiring additional factual verification | `src/content/pages/en/tour.md:88` | “Learner Teen Group Area” | Room description may relate to the Club; no inference that ABA serves teens. Later scoped tour review, unchanged |
| Unresolved and requiring additional factual verification | `src/content/pages/en/client-forms.mdx:82,83` | “Step 4a: Print & Sign (Age 2 - 5)”; “Step 4b: Print & Sign (Ages 6+)” | Diagnostic paperwork bands, not general ABA eligibility; later forms/evaluation review, unchanged |

Context exclusions, deliberately not labeled ABA eligibility errors:
`en/faqs.mdx:34` discusses learners of any age and generic intervention evidence;
`en/early-signs-autism-by-age.mdx` discusses 12–24 months / 2–4 years;
en/es diagnostic-guide tables discuss 2–4 years;
en/es/ar executive-functioning articles discuss developmental stages including
adolescence (`en:82,96,110,112`; corresponding translated sections);
homepage testimonials mention starting at 21 months. No Arabic AIA ABA
eligibility claim matching the superseded range was identified. None of these
files was edited, and this screening is not a clinical or linguistic certification.

### Human-decision / Factual-verification Register

These are held questions for later reconciliation or activation, not reasons
to defer the already-confirmed ABA age correction:

| Area | Evidence / limitation | Current safe disposition and decision needed |
| --- | --- | --- |
| Program availability | Both ABA pages retain 2–6 Academic Readiness/Academy; both Club pages retain 8–17 and schedules; drawer claims enrollment open | Preserve distinct source-supported ranges. Confirm present enrollment, electives and schedules operationally before activating inquiries; do not infer closure or new ages from unrelated Library rewrites |
| Insurance participation | Insurance page shows BCBS, Aetna, Optum, Tricare, AHCCCS and United Healthcare logos. Local drawer specifically lists United Healthcare AHCCCS and additional payment choices; live runtime form fields were not executed/recovered | Logos do not prove all AHCCCS plans or all offered services are covered. Retain existing plan-specific wording; AIA billing must confirm exact networks, service coverage, authorizations, referrals and diagnostic prerequisites. No universal coverage rule added |
| Service geography / waits | Client pages claim no wait list and Phoenix-area coverage; schedule page additionally claims Tucson in-home service; its audience copy demonstrably conflicts with E | No new locations, wait times or staffing claims. Existing geography/wait copy held pending operational confirmation; inquiry drawer's guarantee language is also unverified, not strengthened |
| Clinical/evaluation claims | Services pages retain “1 in 36 … each year”; evaluations describe multidisciplinary teams and ADOS for children of all ages | Public page text is not independent clinical validation. Clinical owner should verify statistic/annual wording, actual evaluation staffing, accepted patient ages and referral/diagnosis requirements; no clinical replacement was invented |
| Referrals / intake | Production referrals uses a general Jotform; local generic FormShell cannot establish clinician referral fields or delivery workflow. Intake describes packets but has no functioning packet collection | Referral CTA exposes the disabled existing local form; intake goes to the local consultation page. Approve routing, required fields and privacy before backend activation; no successful referral or completed intake is implied |
| CTA/form backend | Production Jotform URLs resolve; local forms explicitly unavailable | Review the intentional local inquiry destinations as part of this milestone. Production collection, callback promises, backend choice and privacy approval remain separately gated; no Ads tags or analytics added |
| Contact freshness | Potentially stale schedule page retains `info@abaclinicaz.com`, whereas Club inquiry uses `info@azinstitute4autism.com` | Different addresses alone do not prove an error. Keep existing source-backed addresses, confirm intended consultation contact with AIA in a scoped follow-up |

### Multilingual Review Register

- **Explicitly approved wording:** Spanish ABA 1:1 and consultation now use
  `18 meses a 8 años`. The ABA introduction replaces `personas ... diagnosticadas`
  with `niños ... diagnosticados`, preserving grammatical agreement. Both
  consultation references remove `y adolescentes`. These implement AIA's
  approved correction, not invented translations. Have a human Spanish editor
  review the revised sentences before publication; fluency is not self-approved.
- **Source-supported proper name:** `Academia Preparatoria AIA©` becomes
  `AIA Preparatory Academy©`, matching Spanish production; no curriculum claim
  or destination changed. Include this branding choice in language review.
- **Existing labels retained:** `Hacer una cita`, `Programar una cita`, `Enroll Now`
  and `Take the First Step` match the corresponding production labels. The last
  two remain English on Spanish pages; do not fabricate translations here.
- **Existing form limitation:** FormShell's explanatory copy/fields are English
  even on the Spanish consultation route. No sourced Spanish form was available
  from the unexecuted embed. Additional form localization belongs to a separately
  supported review (Phase 7/form milestone), not speculative translation now.
- Spanish evaluation body, Club rules, intake steps and existing translations
  otherwise remain intact. The services breadcrumb now correctly links to `/es`.

### Validation and Existing Test-harness Defect

The first `test:markdown` invocation failed in suite setup with
`Route is not published: /school-en`. Phase 4B's FAQ component resolves the
publication manifest from Astro.url, but the older Markdown test rendered real
FAQ-bearing content at synthetic aliases. An isolated copy of the **untouched
4079156 checkpoint** reproduced exactly the same failure. This was not caused
by the Phase 5A content changes.

The bounded correction is in `tools/markdown.test.mjs`: keep test aliases as
lookup keys but render each real record at its actual language/collection URL.
All original 20 assertions remain. No parser, FAQ, publication rule, route file
or application code was changed to make the test pass. No failure was ignored.

`tools/seo.test.mjs` adds four actual-Astro regression tests covering approved
general ABA ages/audiences, retained program ranges/ratios/payment distinctions,
14 CTA/breadcrumb cases with real route/fragment targets, disabled forms,
language/direction, one H1 and the repaired evaluation description. The existing
22 SEO tests and 33 publication tests remain. The old ABA section fragment is
also asserted. Tests use temporary copies; no fixture routes enter `dist`.

| Check | Final result |
| --- | --- |
| `npm run test:seo` | 26 passed; zero failures/skips |
| `npm run test:publication` | 33 passed; zero failures/skips |
| `npm run test:markdown` | 20 passed after the verified baseline fixture correction; zero failures/skips |
| `npm run build` (initial normal) | Passed; 96 HTML routes; zero Astro check errors/warnings/hints |
| `PUBLIC_ALLOW_INDEXING=true npm run build` | Passed; 96 routes, 92 sitemap URLs, 174 hreflang links across 53 pages |
| `npm run audit:routes -- --check` | Passed against newly generated offline evidence in each build mode |
| `npm run audit:links` | Zero broken internal links |
| `npm run audit:blog` | Zero blog content audit failures |
| `npm run audit:images` | Zero missing mapped page images |
| `npm run build` (final normal) | Passed; 96 routes, all `noindex,nofollow`, empty sitemap, no hreflang |
| `git diff --check` | Passed |

Source fingerprint refreshes used `npm run audit:routes -- --offline`, never a
new crawl. The final saved reconciliation is
[`route-reconciliation-2026-10-09-offline.json`](route-reconciliation-2026-10-09-offline.json).
All 70 historical discovery requests, original observation dates, baseline and
summary totals are preserved by deep comparison. Changed local rows reflect
five source-extension changes and the repaired description, not newly discovered
production routes. Historical totals remain 129 discovered route keys, 102
production-sitemap entries, 96 generated local routes, 92 overlaps and ten
historical sitemap routes missing locally. Do not confuse that historical
102-entry inventory with the current local 92-entry production-policy sitemap.
The intermediate indexing reconciliation is retained only under `/tmp`.

All-route generated HTML comparison against the starting staging artifact
confirmed identical route sets, canonicals, document titles, visible H1s,
HTML languages/directions, JSON-LD objects and header/language-switcher links.
All stable pre-existing main-content IDs remain; random contact-obfuscation IDs
are excluded. Visible prose differs only on the six eligibility/heading pages;
other edits change links or the one description without changing their labels.
There are no new images, styles, layout components or client scripts. MD-to-MDX
conversion preserves headings, paragraphs and lists apart from the documented
CTA conversion; existing styled Button markup is reused.

The indexing artifact was checked at **01:34:18 UTC**; the restored staging
artifact was independently checked at **03:33:00 UTC**. The later check time
reflects resumed work, not a new production observation. Production mode still
has 92 `index,follow` routes and the same four `noindex,follow` exceptions.
`/ar` remains unpublished, and all noindex/external-canonical exclusions persist.
All 63 BlogPosting, 21 FAQPage and 96 existing organization blocks are unchanged.
No publication, canonical, schema, robots, sitemap or translation policy changed.

Browser-level desktop/mobile validation could not run because the browser
transport was unavailable. Generated HTML, existing CSS reuse, one-H1/ID checks,
link and image checks support the content review; they do not certify visual
or interactive browser fidelity. No claim of completed pixel matching is made.

### Read-only Structural Review and File Inventory

There are **18 logical changed/new files**, including five `.md` → `.mdx`
conversions (23 paths in an unstaged status listing). All paths below are
relative to `www/`:

| Classification | Files | Why required |
| --- | --- | --- |
| English content | `src/content/pages/en/aba-therapy.mdx`, `autism-evaluations.mdx`, `client-consultation.md`, `learner-social-club.mdx`, `schedule-consultation.mdx` | Confirmed ABA facts, one source heading/description and safe local CTAs |
| English content conversions | `src/content/pages/en/referrals.md` → `.mdx`; `aba-therapy-intake-process.md` → `.mdx` | Use the existing Button component for previously inert CTA labels; referral range correction |
| Spanish content | `src/content/pages/es/client-consultation.md`, `learner-social-club.mdx`, `services.mdx` | Approved audience/age correction, local enrollment destination, localized breadcrumb |
| Spanish content conversions | `src/content/pages/es/aba-therapy.md` → `.mdx`; `autism-evaluations.md` → `.mdx`; `aba-therapy-intake-process.md` → `.mdx` | Approved ages/source branding and existing-label Button CTAs, with stable record identity |
| Regression coverage | `tools/seo.test.mjs`, `tools/markdown.test.mjs` | Four focused service tests; restore original Markdown suite compatibility with accepted Phase 4B |
| Dated evidence, new | `reports/phase-5a-source-evidence-2026-10-09.json` | Cache-aware targeted observations, separate from historical audit |
| Offline reconciliation, new | `reports/route-reconciliation-2026-10-09-offline.json` | Refresh local source evidence without rewriting production observations |
| Milestone report | `reports/migration-summary.md` | This reconciliation, audit, decisions, validation and review |

Reviewed changes do not touch dependencies, application configuration, shared
SEO/publication/translation/schema code, FormShell, ServiceInquiryDrawer,
ServicesPage, shared service data, homepage, Library, team, employee portal,
Ads handoff/landing pages, consent, analytics, deployment or Git configuration.
No routes, H1 fields or translation relationships were added. No forms were
activated, submitted or connected. No files are staged; no commit or push was
made. Phase 5B and all other reconciliation phases remain unstarted.

The existing untracked `merge-plan.md` remains unchanged, SHA-256:
`015db80cdbaf7d68799265d2070db760155c342b432ce6795f17ebc2218c641c`.

## Phase 5B — Homepage Content Reconciliation (2026-10-09 UTC)

Implemented on `faithful-astro-migration`, starting at the accepted Phase 5A
checkpoint `051fb7c1081614329dd65cf9e72a496d851b2c37`. This section records a
new observation and implementation milestone; it does not replace the earlier
audit or Phase 5A evidence. Scope is the English `/` and Spanish `/es` homepages,
their section model/components, focused tests and documentation. Detailed visual
fidelity remains Phase 8. No changes are staged or committed.

### Evidence, Dates and Cache Limitations

Evidence labels used below:

- **E — Implemented from confirmed AIA editorial requirements:** the user's
  organizational confirmation dated **2026-10-08**: general ABA eligibility is
  **18 months through 8 years / 18 meses a 8 años**; general ABA audience wording
  referring to children and teens is superseded. This decision is resolved.
- **P — Implemented from production evidence:** exact content observed in this
  phase's bounded public GET responses. Publication on the retrieved page is
  verified; current origin freshness and underlying business eligibility are
  not independently established.
- **R — Retained because current content is supported:** existing content matches
  those retrieved responses, subject to the same cache limitations.
- **D — Deferred due to uncertainty:** needs factual/operational or language review.
- **O — Outside Phase 5B scope:** preserved for its assigned later milestone.

New machine-readable evidence:
[`phase-5b-source-evidence-2026-10-09.json`](phase-5b-source-evidence-2026-10-09.json).
It records exact URLs, UTC timestamps, status, selected cache headers, body
hashes, source metadata/headings, section order, important copy, lists, images
and alt text, CTA destinations, before/after field changes and output checks.
Raw HTML, cookies and broad production captures are not committed.

| Request | Retrieval UTC, 2026-10-09 | Status / cache evidence |
| --- | --- | --- |
| `https://www.azinstitute4autism.com/robots.txt` | 03:47:42 | 200; HIT; permits these homepage/asset requests; avoids disallowed preview/cache-buster paths |
| `https://www.azinstitute4autism.com/` | 03:48:03.652341 | 200; `X-HS-CF-Cache-Status: HIT`; Last-Modified and X-HS-Prerendered `2026-10-03 23:58:31 GMT` |
| `https://www.azinstitute4autism.com/es` | 03:48:05.022332 | 301 to `https://www.azinstitute4autism.com/es/`; Cache-Control `max-age=120` |
| `https://www.azinstitute4autism.com/es/` | 03:48:06.491271 | 200; HIT; Last-Modified and X-HS-Prerendered `2026-10-04 00:02:11 GMT` |
| `https://www.azinstitute4autism.com/?migration_review=20261009T0348` | 03:48:07.882981 | 200; HIT; byte-identical to first English response |
| `https://www.azinstitute4autism.com/es/?migration_review=20261009T0348` | 03:48:09.218681 | 200; HIT; byte-identical to first Spanish response |
| `https://www.azinstitute4autism.com/hubfs/hsa-fsa-accepted.png` | 03:51:40.337258 | 200 PNG; `CF-Cache-Status: HIT`; Age `1891407`; Last-Modified `2026-07-07 19:44:58 GMT`; ETag recorded in JSON |

Seven successful HTTP requests, including the redirect response and image;
no production writes, script execution, submissions, API calls or broad crawl.
Homepage requests were sequential, at least 1.1 seconds apart, with
`Cache-Control: no-cache, max-age=0` and `Pragma: no-cache`. HTML responses have
`Cache-Control: s-maxage=36000, max-age=5`,
`X-HS-Cache-Control: s-maxage=36000, max-age=0`, `Server: cloudflare` and CF-Ray
identifiers. Age, ETag and CF-Cache-Status were absent on those HTML responses.
The first web-retrieval attempt for robots failed; direct retrieval succeeded.
An independent browser attempt failed with `Transport closed`.

**Fresh retrieval is not proof of fresh origin content.** The English hero
still says “children and teens,” even though its services headings and metadata
have changed. The Spanish hero already says “niños,” but the duplicated
responsive services headings disagree: one still says “niños y adolescentes.”
AIA's E correction governs both local homepages. Query variants and no-cache
headers did not establish origin freshness; retries stopped after one recheck
per language. Other source-backed copy is carried forward with these limitations.

### English / Spanish Section Inventory and Resolution

All existing YAML sections were inspected, including headings, body text,
list order, image filenames/alt text and CTA references. Paragraph/list text
comparison normalized whitespace and typographic apostrophes; existing local
editorial formatting, descriptive alt text and historical quotations are retained.
No existing section was found to have been removed in either retrieved page.

| Order | Section | English comparison / resolution | Spanish comparison / resolution | Classification |
| --- | --- | --- | --- | --- |
| 1 | Hero | Preserve eyebrow, H1 markup, image, geography and CTA; remove “and teens” from body despite stale source | Preserve eyebrow, H1 markup, image, geography and CTA; remove “y adolescentes,” also matching source hero | E; remaining fields R |
| 2 | Services introduction / commitments | Replace old audience heading with ABA-specific approved age range; add source word “Treatment” to “Integrated Therapy”; retain five service items and seven commitment items/order | Replace old audience heading with approved Spanish range; retain five services, seven commitments, images and source-supported existing wording | E, P, R |
| 3 | Benefits | Heading, subheading, four benefits, video URL/title/image supported; retained | Own published wording, four benefits and same video retained | R |
| 4 | Skills | Six skills and icons/order supported; retained | Six existing translated labels/icons/order supported; retained | R |
| 5 | Insurance | Heading/body and six carrier logos supported; retained | Own heading/body and same six logos supported; retained | R, D for coverage verification |
| 6 | ESA | Heading, one paragraph, dollar range and ADE image supported; retained | Own published heading/paragraph and ADE image supported; retained | R, D for eligibility/amounts |
| 7 EN only | HSA/FSA | Missing locally; add exact published heading/paragraph and local copy of image, after ESA and before financial help | No equivalent in retrieved Spanish page; omit entirely | P; no invented translation |
| 8 EN / 7 ES | Financial help | Existing heading, uninsured assistance/pay-over-time paragraph and CTA supported; retained | Own existing heading, paragraph and CTA supported; retained | R, D for program availability |
| 9 EN / 8 ES | Process | Six cards present but inert locally; restore links and two incomplete source labels | Six source labels already match; restore links to Spanish intake headings | P; explicit local destination adaptation |
| 10 EN / 9 ES | Clinical director | Heading, quote, Rula Diab name/credentials/photo/signature supported; unchanged | Own published heading/quote/name/credentials preserved | R |
| 11 EN / 10 ES | Testimonials | Update heading “What Clients Are Saying” → “What Clients Say”; all 11 quotations/authors unchanged | Existing heading “Lo que dicen los clientes” and same 11 published English quotations unchanged | P for EN heading; R for quotes |
| 12 EN / 11 ES | BACB/CASP logos | Missing locally; restore unlinked image row immediately after testimonials | Same row published; restore with exact source alt text, which is English | P |

The logo row is inside production `<main>`, before the global footer, not a
footer redesign. Its images already exist locally. No invented membership or
certification paragraph, link or organization schema accompanies them.

### Exact Substantive Copy and Metadata Changes

All P observations refer to the two homepage timestamps above; E refers to
2026-10-08. Full previous/current field values are also in the evidence JSON.

| Route / field | Previous Astro | Resolution / source |
| --- | --- | --- |
| `/`, title | `ABA Therapy Near Me \| Arizona Institute for Autism \| Scottsdale` | P: `Scottsdale ABA Therapy for Children \| Arizona Institute for Autism` |
| `/`, description | `Arizona Institute for Autism: center for behavioral health & education services located in Scottsdale. We serve individuals with Autism and their families.` | P: `Arizona Institute for Autism: center for behavioral health & education services in Scottsdale. We serve children diagnosed with autism and their families.` |
| `/`, hero body phrase | `clinical care for children and teens who have an autism diagnosis` | E: `clinical care for children who have an autism diagnosis`; remaining sentence/geography untouched |
| `/`, services heading | `ABA Therapy Services for Children and Teens Diagnosed with Autism Spectrum Disorder` | E: `ABA Therapy Services for Children 18 Months Through 8 Years Diagnosed with Autism Spectrum Disorder`; production removes “and Teens,” while the explicit range comes from E |
| `/`, commitment item | `Integrated Therapy` | P: `Integrated Therapy Treatment` |
| `/`, process step 3 | `Fill and Sign an Intake Packet` | P: `Fill and Sign a Client Intake Packet` |
| `/`, process step 4 | `Verify Billing and Insurance` | P: `Verify Billing and Insurance Information and Benefits` |
| `/`, testimonials heading | `What Clients Are Saying` | P: `What Clients Say` |
| `/es`, hero body phrase | `atención clínica experta para niños y adolescentes con diagnóstico de autismo` | E + P: `atención clínica experta para niños con diagnóstico de autismo` |
| `/es`, services heading | `Servicios de terapia ABA para niños y adolescentes diagnosticados con TEA` | E: `Servicios de terapia ABA para niños de 18 meses a 8 años diagnosticados con TEA` |

Both visible H1s are unchanged: English `Behavioral Health & Special Education`
(with its existing line breaks) and Spanish `Salud Mental y Educación Especial`.
Spanish title and description remain its own published values; they were not
translated from the revised English metadata. The age range appears only in
ABA-specific headings, not imposed on evaluations, the Social Club or other
services. Historical testimonials, including the “21 months year old” statement,
are untouched. This resolves the four homepage findings deferred by Phase 5A;
its original eligibility audit remains intact as historical evidence.

### English HSA/FSA Implementation and Source Qualifications

Exact published heading:
**Health Savings Accounts (HSA) & Flexible Spending Accounts (FSA)**.

Exact substantive paragraph (only source nonbreaking whitespace normalized):

> Many families can also utilize their HSA and FSA to fund care, which allow you to pay for qualified clinical services using pre-tax dollars. The Arizona Institute for Autism (AIA) accepts Health Savings Account and Flexible Spending Account cards directly, making it easier to manage out-of-pocket costs for your learner's treatment. If you participate in one of these employer-sponsored or individual accounts, you can seamlessly apply your available funds toward tuition and fees for our ABA programs.

The optional `home.hsaFsa` content block uses `heading`, `body`, `image` and
`imageAlt`, validated through the existing Zod homepage model. Its dedicated
`HomeHsaFsa.astro` component renders after ESA/before financial help, matching
production: image left (5/12), text right (7/12), tinted background, one heading
and one paragraph. A single-column layout applies below 760px. The exact
500×425 source PNG is self-hosted as `public/assets/images/hsa-fsa-accepted.png`
(40,208 bytes), SHA-256
`25d1a0c6ed1e156e0861071fa409b5b6908b7f3b02407066a0d3b7014b25c195`.
Source alt `hsa-fsa-accepted` is preserved. The artwork itself says
“HSA APPROVED FSA”; it is published artwork, not an independent certification.

The qualifiers “Many families,” “qualified clinical services” and “If you
participate” are preserved. The section does not claim every expense is eligible,
provide a payment form, collect card data, load third-party scripts or add any
JSON-LD. Spanish has no `hsaFsa` block and renders no empty placeholder.

`home.logos` is a separate optional `items: [{ file, alt }]` block for the
observed final image row. Both optional sections are also supported by the
existing `sections:` discriminated union. Existing homepage records without
these blocks still render. `home.process.steps[].href` is optional; old steps
without destinations remain plain content. The existing process-card wrappers,
equal-height script, labels and styling remain except for accessible links and
focus indication. README documents all new editing fields.

### CTA Reconciliation

Production link labels/destinations are recorded verbatim in the evidence JSON.
No production form was submitted or activated. The four destination pages were
HTTP 200 in Phase 5A's **2026-10-09 01:26:28–01:26:38 UTC** observations; those
checks are explicitly inherited, not claimed as new Phase 5B retrievals.
Generated local destinations and every new fragment are checked by regression
assertions; all links/images pass local audits.

| CTA / locale | Production destination | Previous Astro | Intended Astro / verification |
| --- | --- | --- | --- |
| EN `Get Started`, `Schedule an Assessment` (commitments), `Request an Appointment` | `https://www.azinstitute4autism.com/client-consultation?hsLang=en` | Same labels, `/client-consultation` | Retain local route; all three links resolve; disabled form remains explicitly unavailable |
| ES `Empezar`, `Programar una evaluación` (commitments), `Solicitar cita` | `https://www.azinstitute4autism.com/client-consultation?hsLang=es` | Same labels, `/es/client-consultation` | Retain real Spanish route rather than copying the production English-path destination; disabled form remains unavailable |
| Six EN process cards | `/aba-therapy-intake-process#first-step` through `#sixth-step` | Plain labels, no links | Add links to existing EN intake heading IDs below; step 3/4 labels corrected as above |
| Six ES process cards | English-path `/aba-therapy-intake-process#first-step` through `#sixth-step`, with Spanish labels | Plain labels, no links | Add links to existing Spanish intake heading IDs below; labels unchanged |
| Learner journey video, both | `https://www.youtube-nocookie.com/embed/EczPH1jx9mc?si=beestOCaO4tL7Re6` | Same existing VideoCard destination | Retained; not played/fetched; privacy/loading behavior unchanged |

Intake destination mapping preserves section intent without editing the
out-of-scope intake records or copying nonexistent local HubSpot fragment IDs:

| Step / production fragment | EN local fragment on `/aba-therapy-intake-process` | ES local fragment on `/es/aba-therapy-intake-process` |
| --- | --- | --- |
| 1 / `first-step` | `1-complete-the-learner-information-form` | `1-complete-el-formulario-de-información-del-estudiante` |
| 2 / `second-step` | `2-speak-with-an-aia-client-advocate` | `2-hable-con-un-defensor-del-cliente-de-aia` |
| 3 / `third-step` | `3-fill-out-and-sign-the-client-intake-packet` | `3-complete-y-firme-el-paquete-de-admisión-del-cliente` |
| 4 / `fourth-step` | `4-verify-insurance-and-billing-information` | `4-verificar-la-información-del-seguro-y-la-facturación` |
| 5 / `fifth-step` | `5-schedule-your-childs-assessment` | `5-programa-la-evaluación-de-tu-hijo` |
| 6 / `sixth-step` | `6-collaborate-on-a-personalized-care-plan` | `6-colaborar-en-un-plan-de-atención-personalizado` |

No new route, redirect or form-activation behavior is introduced. Existing
consultation forms continue to state that online submission is not connected;
backend, privacy and workflow decisions remain separately gated.

### Human-decision and Multilingual Review Registers

| Classification / area | Supported observation | Uncertainty / action |
| --- | --- | --- |
| R + D — Insurance | Both homepages explicitly name BCBS AZ, Aetna, Optum, Tricare, **United Healthcare AHCCCS**, UnitedHealth; six logos match | Retain exact plan-specific wording. AIA billing must confirm products, networks, authorization/diagnosis/referral requirements and covered services; no inference that all AHCCCS or all insurer products are accepted. Carries forward Phase 5A's gate |
| R + D — ESA | Both sources give `$4,000 – $6,500` annually and say eligible families can apply funds to ABA tuition; ADE-vendor wording matches | Preserve existing qualified source wording, not a guarantee. Verify current award amounts, vendor status and eligible programs/expenses with AIA/ADE before public launch; no universal ESA eligibility inferred |
| P + D — HSA/FSA | EN source says AIA accepts account cards and funds may apply toward qualified care and ABA tuition/fees | Exact published copy/art retained. AIA billing/account administrators must verify card acceptance, eligible expenses and any additional qualifications; source alone does not establish reimbursement/tax eligibility. No independent financial advice or promise added |
| R + D — Financial assistance | Both sources describe assistance for the uninsured and paying over time | Eligibility, current availability and terms remain unverified. No financing product, universal approval, interest rate or new payment integration invented |
| R + D — Geography / availability | Existing Phoenix-area city list and in-center/in-home program commitments match retrieved sources | Current staffing, openings and service availability still need operational confirmation under Phase 5A's register; no new claims added |
| D — CTA/backend | Local EN/ES consultation and intake routes exist and resolve | Forms stay disabled. Backend, consent, privacy, callback/packet workflow and activation require separate authorization |
| D — Spanish language review | Hero removes `y adolescentes`; services heading inserts explicitly approved `18 meses a 8 años` | Human Spanish editor should review the two revised sentences in context. This is approved wording applied narrowly, not a fabricated translation or self-approved fluency |
| P + D — Spanish source differences | Spanish metadata/section wording differs; no HSA/FSA section; English testimonials and new logo alt text are published that way | Preserve differences. Human language/accessibility review may later localize source-supported alt text/quotes; no translations invented here |
| O — Detailed fidelity / unrelated content | HSA and logos get scoped responsive CSS; other component styling retained | Phase 8 full-page desktop/mobile comparison, spacing/crops/typography and any shared header/footer reconciliation remain outstanding; Library, services beyond 5A, employee portal and Ads work untouched |

There is **no unresolved human decision about general ABA ages or the removal
of the superseded audience wording**. Remaining questions above concern separate
financial, operational, language or integration facts.

### Validation and Reproduction

Run from `www/`:

```sh
npm run test:seo
npm run test:publication
npm run test:markdown
npm run build
PUBLIC_ALLOW_INDEXING=true npm run build
# Refresh local fingerprint using saved evidence, without any live crawl:
npm run audit:routes -- --offline --evidence reports/route-reconciliation-2026-10-09-offline.json
npm run audit:routes -- --check
npm run audit:links
npm run audit:images
npm run audit:blog
npm run build
# Preserve a separate reconciliation of the restored staging artifact:
npm run audit:routes -- --offline --evidence reports/route-reconciliation-2026-10-09-offline.json
npm run audit:routes -- --check
git diff --check
```

The tool creates a new timestamp-suffixed report rather than overwriting existing
evidence. This phase retains the final staging reconciliation as
[`route-reconciliation-2026-10-09-offline-03-57-29-199Z.json`](route-reconciliation-2026-10-09-offline-03-57-29-199Z.json).
The intermediate indexing reconciliation was moved to `/tmp` after its check.
All **70 inherited production requests**, observation timestamps, production
route records, original baseline and summary totals were preserved by deep
comparison. No new production crawl was used to refresh the fingerprint.
Only the local `/` route record changes title/description; generated route keys
are unchanged. Historical discovery still has 129 route keys and 102 source
sitemap URLs; those are not the current local 92-entry production-policy sitemap.

| Check | Result |
| --- | --- |
| `test:seo` | **30 passed**, including four new homepage tests and all earlier SEO/service cases |
| `test:publication` | **33 passed** |
| `test:markdown` | **20 passed** |
| Initial normal, indexing-enabled, final normal builds | All pass; each **96 routes**, Astro check **0 errors / 0 warnings / 0 hints** |
| `audit:routes -- --check` | Pass for indexing and restored staging evidence; offline, no network |
| `audit:links` | **0 broken internal links** |
| `audit:images` | **0 missing mapped page images**; focused test also checks every homepage `<img>` asset |
| `audit:blog` | **0 failures** |
| `git diff --check` | Pass |
| Browser | Unavailable: `Transport closed`; no browser-level visual/interactive verification claimed |

The first SEO run found two incorrect assumptions in the new tests: an Astro
script/style node intervenes after testimonials, and the process and commitment
CTAs share an assessment label. Assertions now check the next content element
and distinguish consultation links from process links. Rerun passed all 30 tests;
no application defect or relaxed route/eligibility requirement was involved.

Four focused tests cover valid/incomplete homepage schema input, present/absent
optional blocks, old unlinked steps, approved ABA ranges and absent superseded
audiences, retained 11 historical testimonials, source metadata, section order,
English-only HSA/FSA and its qualifications, final logo row, one H1, same-language
CTA routes and all 12 real intake IDs, disabled destination forms, local image
files and absence of financial JSON-LD. Optional blocks/links are removed only
inside the existing temporary fixture build to verify backward compatibility.
Existing tests exercise publication metadata in both modes without duplicating
its policy implementation.

| Generated-output invariant | Indexing-enabled, checked 03:56:49 UTC | Final staging, checked 03:57:28 UTC |
| --- | --- | --- |
| HTML routes | **96**, identical route set | **96**, identical route set |
| Sitemap URLs | **92** | **0** |
| Hreflang | **174 links / 53 pages** | **0 links** |
| Robots | Existing 92 indexable / four noindex exceptions | Every page **`noindex,nofollow`** |
| Canonicals, languages/directions, visible H1s | Unchanged across all routes | Unchanged across all routes |
| JSON-LD | Unchanged: 96 MedicalOrganization, 63 BlogPosting, 21 FAQPage | Same objects/counts |

All-route comparison against the starting staging artifact confirmed unchanged
header/footer/language-switcher destinations and unrelated main text; scripts
and styles are excluded from visible-text comparison (contact-obfuscation IDs
are randomized). Only `/` and `/es` change body copy. Only `/` changes title,
description and their existing social-metadata projections. No canonical,
indexing, translation, x-default, sitemap, display-H1, social-metadata or
structured-data policy implementation changed. `/ar` stays unpublished; all
four noindex/external-canonical exceptions remain excluded as before.

### Read-only Structural Diff Review and Complete File Inventory

**13 files: eight modified, five new.** Paths below are relative to `www/`.

| Classification | File | Reason |
| --- | --- | --- |
| English content | `src/content/pages/en/index.md` | Approved audience/range, source metadata/labels, HSA/FSA, intake links and logo data |
| Spanish content | `src/content/pages/es/index.md` | Two approved audience/range edits, existing-label Spanish intake links and source logo data; no HSA/FSA |
| Content model | `src/types/home-sections.ts` | Optional HSA/FSA and logo blocks plus optional process destinations; no Schema.org additions |
| Homepage composition | `src/components/pages/HomePage.astro` | Render optional sections in observed order, including legacy sections-array support |
| Homepage presentation | `src/components/home/HomeProcess.astro` | Link existing process-card content to validated local intake sections; preserve fallback and client script |
| Homepage presentation, **new** | `src/components/home/HomeHsaFsa.astro` | Source-oriented image/text layout with narrow responsive CSS |
| Homepage presentation, **new** | `src/components/home/HomeLogos.astro` | Optional source image row after testimonials, responsive and unlinked |
| Source asset, **new** | `public/assets/images/hsa-fsa-accepted.png` | Exact self-hosted published PNG; provenance/hash above |
| Regression tests | `tools/seo.test.mjs` | Four homepage tests within existing temporary Astro build suite; prior tests preserved |
| Editing documentation | `README.md` | New optional homepage fields and valid locale/fragment destinations |
| Milestone report | `reports/migration-summary.md` | This dated section; earlier history untouched |
| Source evidence, **new** | `reports/phase-5b-source-evidence-2026-10-09.json` | Cache-aware read-only observations, exact changes and verification |
| Offline audit, **new** | `reports/route-reconciliation-2026-10-09-offline-03-57-29-199Z.json` | Updated local fingerprint/metadata; inherited production evidence unchanged |

Reviewed scope excludes services/intake source records, Library, team, employee
portal, separate Ads landing pages, global navigation/footer, deployment,
dependencies/lockfile, Git configuration, form components/backends, analytics,
consent and payment integration. No new translation relationships or fabricated
Spanish sections were introduced. No staging, commit, push or Phase 5C work.

`merge-plan.md` remains untracked and unchanged, verified SHA-256:
`015db80cdbaf7d68799265d2070db760155c342b432ce6795f17ebc2218c641c`.

Tracked-file diff summary (`git diff --stat`; five new files are listed above
and are not included by Git until staged):

```text
 www/README.md                             |   7 +
 www/reports/migration-summary.md          | 324 ++++++++++++++++++++++++++++++
 www/src/components/home/HomeProcess.astro |  17 +-
 www/src/components/pages/HomePage.astro   |   8 +
 www/src/content/pages/en/index.md         |  37 +++-
 www/src/content/pages/es/index.md         |  18 +-
 www/src/types/home-sections.ts            |  14 ++
 www/tools/seo.test.mjs                    | 114 +++++++++++
 8 files changed, 526 insertions(+), 13 deletions(-)
```