SEO Local Skill for Claude

> Example output: examples/seo-local-sweetgreen-com-20260514/LOCAL-SEO-REPORT.md

Local SEO

Score a local business's website against the signals that drive local-pack and "near me" visibility — GBP integration on the page, NAP consistency, on-page local intent, citation footprint on Tier-1 directories, review-platform presence, and local-pack rank for the business's primary keywords. Deliverable is one prioritised fix list, anchored in observable signals.

> Adapted from AgriciDaniel/claude-seo's seo-local skill (MIT). Concept and dimension structure mirror the upstream; backend rewired to SE Ranking + Firecrawl + Google APIs. DataForSEO Maps geo-grid and Business Listings checks from the upstream are dropped (no equivalent backend) — see "Limitations" in the deliverable.

Prerequisites

• SE Ranking MCP server connected (used for local-pack rank, on-page audit data, domain context).
• Claude's WebFetch tool available (used for sense-check fallback when Firecrawl is unavailable).
• User provides: (a) a target domain or homepage URL, (b) at least one primary local keyword (e.g. "dentist Brooklyn", "plumber near me"), (c) target country and ideally city/region for local-pack scoping. Optional: GBP listing URL, Yelp/Trustpilot URLs for review scraping.

Process

• Validate target & preflight. See skills/seo-firecrawl/references/preflight.md for the canonical 3-stage preflight (credit balance, Firecrawl availability, Google APIs). Skill-specific notes:
• Normalise the target (strip protocol from domain; confirm homepage is fetchable). Confirm at least one local keyword was provided — if none, infer from <title> + <h1> of the homepage; if still ambiguous, ask the user before continuing.
• Estimated SE Ranking cost for this skill: ~15–25 credits (1 audit re-check + 3–5 SERP queries + 1 domain overview).
• Firecrawl: optional with WebFetch fallback, ~6–9 Firecrawl credits if available (hard cap 12). When available, steps 4 (GBP-on-page audit), 5 (NAP extraction), and 7 (review scraping) run on the homepage + 5 sample pages + provided review URLs. Without Firecrawl those steps degrade to WebFetch-only — schema/JSON-LD detection and tel: / address element extraction become best-effort prose inspection. Pass --no-firecrawl to force WebFetch-only.
• Google APIs: tier 1 (GSC) unlocks step 8b (GSC local query performance) after the local-pack rank step; tier 2 (GA4) additionally unlocks step 8c (GA4 organic-by-landing-page enrichment). See skills/seo-google/references/cross-skill-integration.md for the full enrichment contract.

• Business-type detection
• Read homepage + /contact + footer prose (WebFetch markdown is enough for this).
• Classify as one of:
• Brick-and-Mortar — visible street address, "Visit us at", embedded Maps iframe.
• Service Area Business (SAB) — no street address, "serving {region}", "we come to you", areaServed in schema without address.streetAddress.
• Hybrid — both signals present (e.g. showroom + service area).
• This determines which checks apply downstream. SAB skips embedded-map and physical-address consistency. Record in LOCAL-SEO-REPORT.md "Snapshot".

• Industry-vertical detection
• From URL patterns (/menu, /practice-areas, /listings, /inventory), <title>, page prose, infer one of: Restaurant / Healthcare / Legal / Home Services / Real Estate / Automotive / Generic.
• This routes citation-source recommendations and schema-subtype recommendations later — load references/local-citation-sources.md for the vertical's Tier-1 directories.

• GBP signals on the page mcp__firecrawl-mcp__firecrawl_scrape (with formats: ["rawHtml"])
• Scrape homepage + /contact (or whichever page has the most local intent).
• From rawHtml extract:
• Embedded Google Maps iframe (<iframe src="https://www.google.com/maps/embed?...">) — record place ID if present.
• Reviews widget / GBP rich snippet markup.
• aggregateRating JSON-LD block (presence is the strongest signal that the site wants stars in SERPs).
• Business hours visibility on page (open-at-search-time correlates with rank — Whitespark's #5 factor).
• Click-to-call: count of <a href="tel:..."> elements.
• GBP profile link: any <a href> to https://g.page/... or https://maps.app.goo.gl/... or https://www.google.com/maps/place/....
• If Firecrawl unavailable: WebFetch markdown can detect a tel: link in some renderings but loses iframes and JSON-LD. Mark Maps embed / aggregateRating / GBP profile-link detection as (skipped — Firecrawl required).

• NAP consistency mcp__firecrawl-mcp__firecrawl_scrape on homepage + 5 sample pages
• Sample pages: homepage, /contact, /about, plus 2 service or location pages (pick from sitemap or top traffic pages).
• For each, extract:
• Visible NAP from rendered prose. Address pattern (street + city + region + postal), phone (tel: href + display format), business name (logo alt, footer, schema name).
• NAP from JSON-LD. Parse every <script type="application/ld+json"> block. Pull name, address.streetAddress, address.addressLocality, address.addressRegion, address.postalCode, telephone.
• Compare across the 6 page samples + schema. Any divergence (different phone format on the contact page vs homepage; "Suite 200" missing from one footer; schema phone in international format while page shows local format) → record in nap-inconsistencies.csv.
• Brick-and-mortar only: if a Maps iframe is present, attempt to read the embedded address from the iframe URL (the place ID and address are URL-encoded). Compare to page/schema NAP. SAB skips this.
• If nap-inconsistencies.csv is empty after the scan, write nap-inconsistencies.csv as a one-line file with header only and note "NAP consistent across {n} pages and schema" in LOCAL-SEO-REPORT.md.

• Local-pack rank tracking DATA_getSerpResults with country/region filters
• For each user-provided local keyword (or the 1–3 inferred from homepage):
• Call DATA_getSerpResults with the user's country and the most specific region/city the API supports (use DATA_getSerpLocations first to confirm a valid location code if the user supplied a city).
• Capture: top 10 organic, local-pack presence (yes/no), the 3 businesses in the local pack if shown (name, rating, review count), AIO presence.
• Cross-check: is the target domain in the top 10 organic? Is the target business name in the local pack?
• Save the parsed result per keyword to local-keywords.csv (columns: keyword,country,location,local_pack_present,target_in_pack,target_pack_position,target_organic_position,top_pack_competitor_1,top_pack_competitor_2,top_pack_competitor_3).
• Note the local-pack-ads caveat: the SE Ranking SERP returns the AI/ads-modified pack as Google serves it. If the local pack shows ads, record that — local-pack ad density jumped from 1% to 22% of mobile US local searches in 2025–2026 per Sterling Sky.

• Reviews scraping mcp__firecrawl-mcp__firecrawl_scrape on user-provided review URLs
• Inputs (user-provided, optional). GBP listing URL (https://www.google.com/maps/place/...), Yelp business URL, Trustpilot business URL, BBB profile URL.
• For each provided URL: scrape with formats: ["rawHtml"]. From the parsed DOM, extract: total review count, average rating, date of most recent review (review velocity proxy), count of owner responses on the most recent 10 reviews.
• Aggregate signals:
• Velocity: ≥1 new review in last 18 days = healthy (Sterling Sky 18-day rule). >21 days since last = "review cliff" risk.
• Volume: <10 Google reviews flags below the magic threshold.
• Star rating: 4.5+ matches consumer filtering thresholds (BrightLocal: 31% only consider 4.5+).
• Owner-response rate on Google: <50% on recent 10 = engagement gap.
• If user provides no review URLs: skip step 7 entirely. Note in LOCAL-SEO-REPORT.md: "Review platforms: not provided. To audit review health, re-run with --reviews 'gbp_url,yelp_url,trustpilot_url'." Don't try to discover them — review-URL discovery is a different problem (and the Maps API path is the one we don't have).

• On-page local-SEO audit DATA_getAuditReport (existing audit) + DATA_getIssuesByUrl on the homepage
• Reuse the existing site audit if one is recent (<30 days, see seo-technical-audit). Don't create a new audit just for local — the audit data already covers title-tag issues, missing schema, mobile usability, etc.
• From the audit, surface the issues that bear on local SEO specifically:
• Title / H1 missing primary city or service term.
• Missing or invalid LocalBusiness JSON-LD.
• Mobile usability issues (mobile = where "near me" happens).
• Schema validation errors (broken aggregateRating, malformed address).
• Defer schema fixes to seo-schema. This skill does NOT generate JSON-LD. If LocalBusiness schema is missing or broken, the deliverable says "Run seo-schema for paste-ready LocalBusiness markup with the correct industry subtype" — that's seo-schema's job and reimplementing it here would duplicate work.

8b. GSC local query performance *(only if google-api.json is present, tier ≥ 1)*

• Pull GSC search analytics for the target property, last 28 days, dimension=query, filtered to local-intent patterns:

python3 scripts/gsc_query.py --property "{config.default_property}" --days 28 --json

• Client-side filter the queries for: contains near me, contains a city/region known for the business, or ends in a place-name. Surface top 10 by impressions.
• If a city-bearing query has impressions >100 and average position >10, that's a local-pack reach gap — flag in LOCAL-SEO-REPORT.md "Top fixes" with the GSC numbers as supporting evidence.
• If property not verified for this account: surface "GSC: {target_domain} not verified — add it in Search Console" and continue.
• See skills/seo-google/references/cross-skill-integration.md for failure modes.

8c. GA4 organic by landing page *(only if google-api.json is present, tier ≥ 2)*

• Pull GA4 top organic landing pages, last 28 days:

python3 scripts/ga4_report.py --report top-pages --days 28 --json

• For multi-location sites, surface per-location-page sessions. If one location page captures 80%+ of organic traffic while peer location pages capture <5%, that's location-page quality variance worth flagging (probable doorway-page or thin-content risk on the underperformers).
• Single-location sites: just record the homepage's organic sessions as one row in the snapshot.

• Citation-presence sample (best-effort) WebSearch (no API key cost)
• For each Tier-1 directory in the vertical's list (load references/local-citation-sources.md), check whether the business has a listing using site:{directory} "{business_name}" queries via WebSearch.
• Cap at 8 directories (Google, Yelp, Facebook, BBB, Apple Maps, Bing Places, plus 2 vertical-specific). Anything beyond is diminishing returns and the user can run their own audit.
• Record in LOCAL-SEO-REPORT.md "Citations" section: detected / not detected per directory, plus the URL of the listing if found.
• Caveat to surface: WebSearch hits are a *sample*, not a comprehensive audit. A "not detected" doesn't prove absence — it proves the listing didn't surface for that specific query. Recommend a paid citation-audit tool (Whitespark, BrightLocal, Yext) for definitive coverage.

10. Synthesise LOCAL-SEO-REPORT.md

• Score the 5 local dimensions on the rubric below, list top fixes (Critical / High / Medium / Low), record limitations.
• Apply the verdict heuristic — see Tips.

Output format

Create a folder seo-local-{domain-slug}-{YYYYMMDD}/ with:

seo-local-{domain-slug}-{YYYYMMDD}/
├── LOCAL-SEO-REPORT.md         (PRIMARY: verdict, scores, top fixes, limitations)
├── local-keywords.csv          (load-bearing: per-keyword local-pack + organic positions)
├── nap-inconsistencies.csv     (load-bearing: only emitted if discrepancies found)
└── evidence/
    ├── 01-homepage-snapshot.md     (Firecrawl raw HTML extracts: NAP, schema, GBP signals)
    ├── 02-nap-page-samples.md      (per-page NAP extracts across 5 sample URLs)
    ├── 03-serp-context.md          (raw DATA_getSerpResults per keyword)
    ├── 04-reviews.md               (per-platform review-page snapshots, only if user provided URLs)
    └── 05-citation-sample.md       (raw WebSearch results per directory check)

LOCAL-SEO-REPORT.md follows this shape:

# Local SEO Report: {domain}

> Snapshot dated {YYYY-MM-DD} · Country: {country} · Region: {region} · Primary keyword: "{keyword}"

## Snapshot
- Business type: {Brick-and-Mortar | SAB | Hybrid}
- Industry vertical: {Restaurant | Healthcare | Legal | Home Services | Real Estate | Automotive | Generic}
- Pages sampled for NAP: {n}
- Local keywords tracked: {n}
- Local pack present on {n}/{m} keywords; target in pack on {p}/{m}
- Review platforms audited: {Google, Yelp, ... | not provided}
- GSC last 28d local-intent queries: {n} queries / {clicks} clicks / {impressions} impressions  *(or `not configured`)*

## Verdict: {STRONG | NEEDS WORK | WEAK}

{One-sentence summary anchored in dimension scores below}

## Dimension scores (0–10)

| Dimension | Score | Top finding |
|---|---|---|
| GBP integration on page | {n}/10 | {one-line} |
| NAP consistency | {n}/10 | {one-line} |
| Local on-page (title/H1/contact/service pages) | {n}/10 | {one-line} |
| Local-pack rank | {n}/10 | {one-line} |
| Reviews & citations | {n}/10 | {one-line} |
| **Composite** | {n}/10 | — |

## Top fixes

### Critical
1. {Specific fix anchored in a finding above. Example: "NAP discrepancy: footer shows '(212) 555-1234' but JSON-LD shows '+1-212-555-9999'. Pick one canonical phone, fix the wrong one." Cite the page/schema source.}

### High
- {fix}

### Medium
- {fix}

### Low
- {fix}

## Local-pack rank summary
- "{keyword 1}": local pack {present/absent}, target {in pack at #n / not in pack}, organic position {n}.
- "{keyword 2}": …
- (Full data: `local-keywords.csv`)

## Reviews health (if audited)
- Google: {n} reviews, {rating} avg, last review {n} days ago, owner response rate {p}%.
- Yelp: …
- Trustpilot: …

## Citation sample
- Google Business: {detected / not detected via site:google.com "..." search}
- Yelp: …
- Facebook: …
- BBB: …
- Apple Business Connect: …
- Bing Places: …
- {vertical-specific 1}: …
- {vertical-specific 2}: …

(Caveat: this is a sample, not a comprehensive citation audit. For definitive coverage, use Whitespark / BrightLocal / Yext.)

## Schema status
- LocalBusiness JSON-LD: {present and valid / present but missing recommended properties / invalid / absent}
- Recommended next step: run `seo-schema {homepage_url}` for paste-ready LocalBusiness markup with the correct industry subtype.

## Limitations
This skill could NOT assess:
- **Geo-grid local-pack rank by lat/long.** Requires a Maps API (e.g. DataForSEO Maps geo-grid endpoint) we don't have. Workaround: pay for Local Falcon, GMB Crush, or BrightLocal Local Search Grid.
- **Comprehensive citation audit.** WebSearch sampling covers ~8 directories; full audits cover 50+. Use Whitespark, BrightLocal, or Yext.
- **GBP Insights data.** Requires GBP API access scoped to the listing owner. Ask the listing owner to export Insights and share.
- **Real-time local-pack rank tracking over time.** This skill is a snapshot. Use SE Ranking's project-level rank tracker (`PROJECT_runPositionCheck`) or pair with `seo-drift` for diff snapshots.
- **DataForSEO Business Listings.** Not in our backend; if the user needs it, they'd need to subscribe to DataForSEO directly.

local-keywords.csv columns: keyword,country,location,local_pack_present,target_in_pack,target_pack_position,target_organic_position,top_pack_competitor_1,top_pack_competitor_2,top_pack_competitor_3,aio_present

nap-inconsistencies.csv columns: source,name,address,phone,page_or_schema_path,canonical_value,divergence_note

Tips

• Respect SE Ranking Data API rate limit: 10 requests per second. Pace the per-keyword DATA_getSerpResults calls sequentially.
• Call DATA_getCreditBalance before running. ~15–25 SE Ranking credits typical, plus 6–12 Firecrawl credits when Firecrawl is installed.
• Verdict heuristic:
• STRONG: composite ≥7/10, NAP consistent across all sampled pages, target in local pack on majority of keywords, valid LocalBusiness schema with industry-correct subtype, ≥10 Google reviews with healthy velocity.
• NEEDS WORK: composite 4–6.9/10, OR 1+ NAP discrepancy, OR target out of local pack on majority of keywords. The "Top fixes" section is the deliverable here — most local-SEO audits land in this bucket.
• WEAK: composite <4/10, OR no LocalBusiness schema and no NAP visible on page, OR target absent from local pack on every tracked keyword. Substantial work required across multiple dimensions.
• Don't generate LocalBusiness schema in this skill. Always defer to seo-schema for that — it has the rich-results validation and industry-subtype routing this skill doesn't replicate.
• Don't generate review URLs from search results. If the user didn't provide a Yelp/Trustpilot/BBB URL, skip review scraping and tell the user to provide URLs in a re-run. Discovering review URLs from a domain is unreliable.
• For multi-location sites with >5 locations, audit one location page per region rather than one per location — the local audit pattern repeats per location, so a sample establishes the baseline. If location-page quality variance is the suspected issue, pair with seo-content-audit on a sample of location pages.
• AI-search local context (ChatGPT, Perplexity, AI Overviews) is not this skill's job — pair with seo-geo (URL-level GEO) or seo-ai-search-share-of-voice (domain-level brand visibility) for AI-search local visibility.
• The 18-day review velocity rule (Sterling Sky) is the most actionable single number from review-platform analysis. If the most recent review is >21 days old, that's a leading indicator of upcoming local-pack rank drop — flag as Critical regardless of star rating.
• Citation directories per vertical: load references/local-citation-sources.md. The list is curated to the directories that move the needle (Tier 1 + vertical-specific), not the long tail.
• Pair with seo-technical-audit if site-wide technical-SEO issues surface in step 8 — this skill scopes to local-relevant findings, not the full audit.