wp-audit-local-standards · git:20261006.ff44a3b · 2026-10-06 · sha256 52c86e69efeb3a2a
wp-audit-local-standards git:20261006.ff44a3bA
Immutable. This exact content never changes and is served at /api/v1/blob/52c86e69efeb3a2a.
--- name: wp-audit-local-standards description: Local SEO audit reference — business-type and vertical detection, NAP consistency sources, LocalBusiness subtype selection, location-page quality gates and citation tiers, with the WordPress option and meta keys each check reads. Use when the wp-audit-seo agent audits a business with a physical location or service area. user-invocable: false --- # Local SEO Audit Standards The `SEO-055` through `SEO-063` checks in `agents/wp-audit-seo.md` are the *what*. This file is the *how* and the *why*: which signals decide a business type, which schema subtype a vertical requires, and where a WordPress site actually stores each value. Local SEO differs from the rest of the SEO audit in one structural way: **most of the ranking surface is off-site** — the business profile, the reviews, the directory listings. A code-only audit cannot see any of it. Everything below is scoped to what a theme, its options and its rendered head can prove, and the checks say so explicitly rather than reporting an absence as a pass. --- ## Applicability Gate Run the local checks only when the site is a local business. Otherwise every check reports `not_applicable` and none of them count toward the score. A site is local when **any two** of these hold: - A `LocalBusiness` node (or a subtype of it) exists in the rendered JSON-LD. - A street address appears in the footer, the contact template or an options-page field. - A `tel:` link exists outside a generic "call us" placeholder. - Rank Math's local SEO module is active (`rank_math_modules` contains `local-seo`). - The theme registers a location post type or an ACF location field group. One signal alone is not enough: a `tel:` link in a footer is as common on a SaaS brochure site as on a dentist's site. --- ## Business Type Detection The type decides which checks apply. Getting it wrong produces false criticals — a service-area business has no street address *by design*. | Type | Signals | Checks that do NOT apply | |------|---------|--------------------------| | **Brick-and-mortar** | Street address in content, footer or schema; map embed with a pin; "visit us", "located at" | — (full set applies) | | **Service-area (SAB)** | No visible street address; "serving <region>", "we come to you", "mobile service"; `areaServed` present without `address.streetAddress` | the map half of SEO-056; the address half of SEO-057 (parity and absence) | | **Hybrid** | Both a physical address and service-area language | — (full set applies, plus `areaServed`) | When signals are contradictory, report the type as `undetermined`, run only the checks that are type-independent (SEO-055, SEO-058, SEO-060, SEO-061, SEO-063) and say so in the report. SEO-059 and SEO-062 still run wherever their own preconditions hold — a `LocalBusiness` node to read `geo` from, more than one location page. --- ## Industry Vertical Detection The vertical decides the required schema subtype. Detect from template and content signals; never from the client's name. | Vertical | Detection signals | Required subtype | |----------|-------------------|------------------| | Restaurant | menu template or post type, dish names, reservations, "dine-in", "takeout" | `Restaurant` | | Healthcare | "patients", appointments, insurance accepted, practitioner bios, HIPAA notice | `MedicalClinic`, `Dentist` or `Hospital` | | Legal | "attorney", practice areas, bar admission, case results, "free consultation" | `LegalService` | | Home services | service-area language, "free estimate", licensed/insured/bonded, 24/7 | subtype + `areaServed` | | Real estate | listings, MLS references, agent bios, brokerage, "open house" | `RealEstateAgent` | | Automotive | inventory, VIN, test drive, service department, "new/used/certified" | `AutoDealer` | No vertical detected → generic `LocalBusiness` is correct, not a finding. **Deprecated subtypes that still appear in older themes:** `Attorney` (use `LegalService`), `VehicleListing` as a business type (use `AutoDealer`), bare `MedicalBusiness` for a clinic (use the specific subtype). Flag these under SEO-058. --- ## Where WordPress Stores Each Value A local audit reads three sources for what is conceptually one fact — the rendered page, the stored settings and the theme's templates — and the discrepancies between them are the finding. Stored settings are two different stores, so they get two columns. | Value | Rendered | Rank Math option (`rank-math-options-titles`) | Options-page ACF field (starter names) | Theme template | |-------|----------|-----------------------------------------------|----------------------------------------|----------------| | Business name | JSON-LD `name` | `knowledgegraph_name` (and WordPress's `blogname`) | — | header, footer | | Address | JSON-LD `address.*` | `local_address` | `business_address` | footer template part, contact template | | Phone | JSON-LD `telephone`; `tel:` href | `phone_numbers` | `contact_phone`, `header_phone` | footer template part | | Opening hours | JSON-LD `openingHoursSpecification` | `opening_hours` | the theme's hours repeater, when it has one | — | | Geo coordinates | JSON-LD `geo.latitude` / `geo.longitude` | `geo` | the theme's map fields, when it has them | map embed attributes | Rank Math keeps its local values inside the `rank-math-options-titles` option, not in an option of their own. An adopted site's field names are its own: read them from its field groups rather than assuming the starter's. The agent reads every option once, with `$WP`, in step 3 of its local procedure — never by parsing PHP. **Bilingual sites.** Options-page fields carry their `_<lang>` suffixes under *both* i18n strategies — `business_address_es` — because options are global and Polylang's one-post-per-language model does not reach them. So always compare every `_<lang>` variant, whichever strategy the project records. An address that is correct in one language and stale in the other is a real NAP discrepancy, not a translation artifact. --- ## NAP Consistency NAP is Name, Address, Phone. The check compares the three sources above pairwise and reports each disagreement separately — a phone that differs between schema and footer is a different defect from an address that differs between two languages. Normalize before comparing, or the check reports noise: - Phone: keep the digits only, then drop, in order, a leading `00`, the site's country calling code (from `addressCountry`) and a leading trunk `0`; compare what is left. The country code is the step that matters — digits alone leave `34900000000` and `900000000` unequal, which is the false positive normalization exists to prevent. - Address: collapse whitespace, lowercase, strip trailing punctuation; treat common abbreviations as equal (`St.` / `Street`, `Ave` / `Avenue`). - Name: strip legal suffixes (`S.L.`, `Inc.`, `Ltd.`) before comparing; report a bare suffix difference as INFO, not WARNING. | Source A | Source B | Normalized | Verdict | |----------|----------|------------|---------| | `+34 900 00 00 00` (JSON-LD, Spain) | `900000000` (footer `tel:`) | `900000000` and `900000000` | same number, no finding | | `+44 20 7946 0000` (JSON-LD, UK) | `020 7946 0000` (footer) | `2079460000` and `2079460000` | same number, no finding | | `+34 900 00 00 00` | `+34 900 00 00 01` | `900000000` and `900000001` | SEO-057 | Absent NAP is more severe than inconsistent NAP: a brick-and-mortar or hybrid business whose address appears nowhere in the rendered HTML fails SEO-057 at WARNING even when the schema is perfect, because the schema alone gives a human visitor nothing. A service-area business has no street address by design, so for it an absent address is not a finding. --- ## Citation Tiers Off-site, so the audit can only detect *references* to these from the site itself — badges, links and `sameAs` entries. Report what is present; never assert that a missing reference means a missing listing. - **Tier 1 (general):** the business profile itself, Yelp, the Better Business Bureau, Facebook, Apple Maps, Bing Places. - **Tier 2 (authority):** chamber of commerce, local press, industry associations. - **Vertical directories:** per-industry listing sites appropriate to the vertical detected above. The `sameAs` array in the Organization or LocalBusiness node is the one citation signal a code-only audit *can* verify, which is why SEO-060 anchors on it. --- ## Location-Page Quality Multi-location sites fail in one characteristic way: templated pages that differ only in the city name. Search engines treat these as doorway pages. **The swap test.** Take two location pages, exchange the city names, and read them. If both still make sense, the pages carry no location-specific content and SEO-062 fires at WARNING. This is a judgment call the agent makes by reading, not a ratio it computes. Quality gates by page count, applied before the audit spends effort: | Location pages | Action | |----------------|--------| | 1–29 | Audit each one | | 30–49 | Audit a sample of 10, report the sample size in the finding | | 50+ | Audit a sample of 10 and report the total at INFO; a full walk is manual-only | **URL structure.** Subdirectories (`/locations/<city>/`) consolidate authority better than subdomains. A store locator whose individual locations have no crawlable URL of their own — rendered entirely client-side — is a CRITICAL finding: those pages do not exist for a crawler. **Schema per location.** Each location page carries its own `LocalBusiness` node with a unique `@id`, linked to the site-wide Organization through `parentOrganization` — schema.org supersedes `branchOf` with it; accept `branchOf` on older markup and never report correct `parentOrganization` markup for lacking it. A single shared `@id` across locations collapses them into one entity. --- ## Schema Properties Required by Google for a local rich result: - `name` - `address` as a `PostalAddress` with `streetAddress`, `addressLocality`, `postalCode` and `addressCountry` Recommended. These are not findings of their own: no check id carries them, and a finding without one can be neither rendered nor tracked across runs. List the absent ones in the local section of the report as recommendations. `geo` is the exception, because SEO-059 checks it. - `telephone`, `url`, `image` - `geo` with `latitude` and `longitude` — **at least five decimal places**; three decimals places the pin roughly a hundred metres off (SEO-059) - `openingHoursSpecification` - `priceRange` — under 100 characters - `areaServed` for service-area and hybrid businesses `aggregateRating` is never on that list. Absent, it is not a finding and not a recommendation, because recommending it is a nudge toward inventing one. Present without real review data behind it, or carrying placeholder values, it is a CRITICAL finding under SEO-063: structured-data spam that risks a manual action. The starter theme must never ship one. Schema is not a ranking factor. It earns rich results and gives AI systems parseable business data. Word the findings that way; a report claiming schema lifts rankings is wrong and the client will eventually be told so. --- ## What This Audit Cannot See Every local report ends with this list. Omitting it implies a completeness the audit does not have: - Business-profile completeness, categories, posts and photos - Review count, rating, velocity and owner responses - Actual local-pack or map position, which varies by the searcher's location - Citation accuracy on third-party directories - Profile insight metrics These need the business-profile account or a paid rank tracker. Say which, so the client knows what the gap costs to close. --- ## Attribution The business-type and vertical taxonomies, the swap test and the citation tiering are adapted from the MIT-licensed `claude-seo` project by AgriciDaniel (<https://github.com/AgricIDaniel/claude-seo>). The WordPress mapping, the i18n rules, the severity assignments and the check codes are this plugin's own.