git:20261006.ff44a3b to git:20261006.b4403b0

+ description: GEO and AI-agent-readiness reference — the ORA and is-agentic scoring model, the GEO-D, GEO-A, GEO-U and GEO-P check catalog, applicability by site type (content, local, merchant, catalog store, SaaS), the AI crawler allowlist for robots.txt and the Content-Signal header, the dynamic llms.txt, ARD catalog (ard.json), agent-skills index, markdown negotiation and other surfaces inc/agentic.php serves, the bin/geo-scan.sh live verifier, and a citability rubric for writing copy that AI engines quote. Use when running /wp-audit --geo, explaining why is-agentic grades a site, fixing GEO findings with wp-agentic-surfaces, choosing which AI crawlers robots.txt allows, or writing section copy for ChatGPT and Perplexity to cite. Not for Rank Math titles, meta descriptions, sitemap or schema configuration (wp-audit-seo-standards), and not for accessibility fixes, which GEO-U01 to U05 only mirror.

+ Scored by **ORA**, surfaced at [is-agentic.com](https://is-agentic.com), in four layers.

+ The live check catalog is `GET https://ora.ai/api/checks`; every count and weight below is a

+ snapshot of it, so re-fetch it before quoting a number as current.

+ This skill informs the audit (`${CLAUDE_PLUGIN_ROOT}/agents/wp-audit-geo.md`) and the fix

+ (`${CLAUDE_PLUGIN_ROOT}/agents/wp-agentic-surfaces.md`), which holds the executable version

+ of every surface. It never runs commands on its own. Terms used throughout: a **crawler** is

+ an AI user agent named in `robots.txt`; the **allowlist** is which crawlers `robots.txt`

+ admits; a **surface** is a URL the theme serves for agents (`/llms.txt`, `/.well-known/…`);

+ the **check catalog** is ORA's list of checks, and the **ARD catalog** is `ard.json`.

+ ## Reference files

+

+ - [references/check-catalog.md](references/check-catalog.md) — every GEO code, the ORA check

+ ids it maps to and whether the theme can fix it. Read when mapping an ORA id to a code,

+ tabulating a code, or deciding whether a finding is fixable or advisory.

+ - [references/surface-templates.md](references/surface-templates.md) — what each surface must

+ contain: the minimum robots body, `llms.txt`, `ard.json`, the agent-skills index,

+ `pricing.md`, markdown negotiation, `Link:` headers, the agent-friendly 404, JSON-LD

+ breadth, trust anchors, `llms-full.txt` and `auth.md`. Read when building or checking a

+ surface.

+ - [references/citability.md](references/citability.md) — the citability rubric and a

+ weak-to-strong rewrite. Read when writing or tightening page copy (`/wp-section`,

+ `/wp-seed`); no audit code scores it.

+

+ advisory either way, because no WooCommerce payment gateway speaks an agent payment

+ protocol (§8).

+ ## 3. Check catalog

+ Codes are `GEO-<layer><nn>`, layer letter D, A, U or P. Each maps to one or more ORA check

+ ids and is the unit the agents tabulate and report; the full table, with what the theme can

+ fix, is [references/check-catalog.md](references/check-catalog.md). GEO-A25 to GEO-A28 are

+ the plugin's own: ORA has no check for them, so they are reported outside the ORA score.

+ and training crawlers below.

+ The minimum robots body is every ALLOW row, then the BLOCK row, after the classic

+ `User-agent: *` block — written out in

+ [references/surface-templates.md](references/surface-templates.md).

+ **The training posture is the owner's decision, and a deliberate one stands.** An existing

+ `Disallow` for a named crawler, or an `ai-train=no` signal, is the site owner's answer, not

+ a defect: the auditor reports it as a deliberate block (consistent or not with the signal,

+ below), and the fixer asks before changing it — it never rewrites a deliberate block back to

+ the default. The default applies only where the site states no posture at all.

+ ## 5. Citability

+ Citability is a property of the **copy**, not the markup: how likely a generative engine is

+ to lift a passage verbatim. The rubric — five weighted components, an optimal passage of

+ 134-167 words, answer-first openings — and a worked rewrite are in

+ [references/citability.md](references/citability.md).

+ ## 6. Surfaces

+ The fixer writes the theme's `inc/agentic.php`, which owns every surface; all are

+ applicability-gated, so a content site emits no API or payment surface. What each must

+ contain is [references/surface-templates.md](references/surface-templates.md). One rule

+ belongs here because both agents act on it: **when Rank Math, Yoast or SEOPress owns the

+ schema, the theme emits no JSON-LD**, so GEO-A07 to GEO-A10 are fixable only when no SEO

+ plugin owns the graph — otherwise `wp-audit-rankmath` fills it.

+ does not parse the JSON.

+ The loop:

+ 1. **Scan** the public host and note the report's scan time.

+ 2. **Map** each failed ORA check id to its GEO code with

+ [references/check-catalog.md](references/check-catalog.md).

+ 3. **Fix** what the theme can fix (`wp-agentic-surfaces`); leave advisory codes as

+ recommendations.

+ 4. **Re-scan once**, after the fix is deployed to the host the scan reads.

+ 5. **Compare.** A fix is resolved **only when its ORA check flips** in a report scanned

+ after the fix — never because the theme file changed. A report whose scan time precedes

+ the fix is the old report: `geo-scan.sh` only starts a scan when none exists, so it

+ returns the pre-fix report until is-agentic re-scans the site. Ask the operator to

+ re-scan at https://is-agentic.com; until then the code is `UNMEASURED`, not resolved.

+ Stop after that one re-scan.

+

+ The evaluator reads a point-in-time scan; a green scan is evidence, not a guarantee.

+

+ - **Payments / AP2 / ACP / UCP / x402 are merchant-only and advisory** for WooCommerce:

+ no WooCommerce gateway implements an agent payment protocol, so the fixer recommends and

+ never fabricates support.