Service-page vocabulary and cross-cutting metadata (CI-14, §12)

[zolizoli]

Stacked on #12 → #11 → #10 → #9. Last PR in the pass. Contains copy written for your marketing pages — please read it before merging. No essay prose was touched and no new CTAs were added (verified: 0 new btn/cal.com elements in the diff).

CI-14 — the vocabulary problem

/services/ai-adoption-audit.html is good writing that matches no query. "AI adoption audit" is an invented category; buyers search "AI readiness assessment" and "AI maturity assessment".

The H1 stays. A named, ownable offering is a real asset, and renaming it to chase a keyword trades something real for something borrowed. Instead:

• A new H2, "How this differs from an AI readiness assessment", which earns the term rather than stuffing it: a readiness assessment scores an organisation before it builds; this audit examines a tool that already shipped and is not being used. Same shape, different evidence.
• Both terms worked into the body naturally (8 × "readiness", 2 × "maturity").
• Meta description rewritten to lead with the term. It previously read "A fixed-scope diagnostic that finds why an internal AI tool has stalled" — accurate, and matching no query. Now 141 chars, within the truncation limit.
• alternateName: ["AI readiness assessment", "AI maturity assessment"] on the Service node.

FAQ

Five questions under "Questions we are asked", each answerable from material already on the page (duration, whether it is a readiness assessment, deliverables, whether we build the fix, whether it suits an unbuilt tool), marked up as FAQPage.

Only questions that genuinely appear on the page are marked up — asserted in the build: all 5 questions and all 5 answers verified present in the rendered visible text.

The og:image finding

The audit reported this page reusing the generic /images/social-preview/home.png. It is worse than that: the page had no Og_image at all, so it inherited the base template default — /site/img/punctuation_spiral.png, an essay illustration. /images/social-preview/ai-adoption.png already existed and was in use on the services card. Now set.

Commissioned Analysis — 408 → 598 words

The thinnest of the three, and it linked to neither sibling while they linked to each other. Two sections added, both grounded in work already published on this site:

• What the tooling actually does — each of the five libraries, what it measures, and why it exists.
• What it looks like finished — the Nagel Index, the Homer essay (21.3% of the Odyssey is open water), the K-Monitor networks (~64,800 articles). Real numbers, all already published here.

All three service pages now link to the other two.

Cross-cutting (§12) — audited, mostly already fine

Audit claim	Finding
Missing alt on founder portraits and the contact image	False. Every <img> in the theme and in about.md has alt. The linter from #12 now enforces it on every page.
/packages.html missing the Terms footer link	False. It is in base.html:150, so every page has it.
og:image reuse	Fixed for the adoption audit. /portfolio.html, /privacy.html and /terms.html do share the homepage image — but no dedicated artwork exists for them, and commissioning some is a design decision, not a lint fix. Reported, not invented.
Title/H1 divergence on four essays	Reported, not changed. The Nagel Index — Does Power Change a Person? against "Does power change a person—or does it reveal them?" is ordinary editorial practice, and the H1 is essay content.
/aporia/epistemic-arcade/ description 169 chars	Was actually 170, and fixed in #12 (a bug in the linter's own quote matching had been truncating it at an apostrophe).
Author bylines on the essays	Not added. A byline means saying who wrote it, and that is not established anywhere in the repo. Four of these essays run 3,000–6,500 words, so this is worth resolving — but with your answer, not a guess.
og:description diverging from meta description on /aporia/magyar-dalszovegek-essze/	Upstream. That page is generated from ~/projects/music_networks; a hand edit here is lost on the next regen.

Schema tidy

All three Service nodes now reference the shared organisation @id instead of describing Crow Intelligence a fourth time. areaServed and audience added where absent.

Verification

Build clean · linter 0 errors, 0 stale exemptions · 57 JSON-LD blocks parse · npm test 20/20 · service pages now 1,268 / 688 / 598 words.

What still needs you

1. Read the service copy. It is written in your voice as best I can read it, but it is your sales surface.
2. Essay bylines — who wrote which of the five.
3. Social images for /portfolio.html, /privacy.html, /terms.html, if you want them distinct.