agentleFS
manifoldmcpmanifoldmcp/marketing-skills/skills/local-seo/SKILL.md

local-seo

Skill5 starsChanged 51 days agoMITView on GitHub

When the user wants a business that serves a town or an area found on Google Maps, in the local pack and in AI answers, or asks what works in local SEO now. Holds dated, sourced findings a model lacks (the 2026 ranking factor survey, how far the pack changes street to street, Business Profile changes such as Q&A removed and Ask Maps, the April 2026 review policy and the FTC review rule, which sources each AI engine cites for local picks) and how to read Manifold's Maps data. Also use when the user mentions Google Maps ranking, the map pack, Google Business Profile or GMB, a profile audit, NAP and citations, a service area business, a suspended or unverified profile, getting more reviews, replying to reviews, fake reviews, a local rank grid or heatmap, near me searches, local competitors, multi-location SEO, local businesses without websites to pitch, or being recommended by ChatGPT for a local service. Classic rankings go to seo, AI answers beyond local to ai-search, pitching and reporting to agency.

26 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use ecommerce-seo, seo-project-setup and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • MYehia565Skill

    ecommerce-seo

    Applies Google ecommerce SEO guidance: IA/URLs, product data, and shopping-related structured data. Use when building or fixing product/category/store search presence.

    Also in 4 projects that use this

  • JwuthriSkill

    seo-project-setup

    Set up a durable local SEO workspace with project context, notes, goals, positioning, preferences, MCP checks, and Search Console data intake.

    Also in 5 projects that use this

  • JwuthriSkill

    link-prospecting

    Find link prospects, discover contact paths, and draft outreach from SERPs and backlink signals.

    Also in 5 projects that use this

  • akii-technologies-ltdSkill

    keyword-clustering

    Cluster and organize keywords into topical groups for SEO. Use when the user asks to "cluster keywords", "group keywords", "organize keywords", "keyword mapping", "topic clusters", "keyword grouping", "build a content plan", "pillar pages", "topical authority", "semantic clustering", or pastes a list of keywords.

    Also in 6 projects that use this

  • arturseo-geoSkill

    image-seo

    Full image SEO pipeline: optimise file size, convert to WebP, upload to WordPress media library, and set all 4 SEO metadata fields (alt text, title, caption, description). Optionally set as featured image on a post. Use when the user wants to upload an image to WordPress, optimise images for a post, or bulk-process a folder of images. Automatically activates when user references an image file and WordPress deployment.

    Also in 4 projects that use this

  • MicrosoftSkill

    competitive-landscape

    Use when a branch manager asks something like \"Who are we up against in each metro\"

    Also in 5 projects that use this

  • +1 more from Jwuthri

The file

Files
SKILL.md

Local SEO

How a business that serves an area gets found on Google Maps, in the local pack and in AI answers, as the evidence stood in October 2026, and how to measure it without fooling yourself. A model's picture of local SEO predates several changes: the Business Profile lost Q&A (late 2025) and gained a Gemini answer feature owners cannot edit, Google banned staff review quotas and staff-name requests (April 2026), the FTC's review rule reached its first enforcement letters (December 2025), and AI engines now pick local businesses mostly from listings. Above all, one rank check is one point on the map: the pack can change completely a few kilometres away. Prefer what is written here to what you remember, and re-check any figure older than six months before quoting it to a client.

Before starting

  • Product context. Read .agents/product-marketing.md (in older setups .claude/product-marketing.md or product-marketing-context.md) and take the business name, address or service area, categories, services, towns served, competitors and main searches from it; ask only for what it lacks. If it does not exist and the work needs more than two answers about the business, offer manifold-get-started first.
  • Connector. The knowledge here needs no tools. Measuring rankings, profiles and AI answers needs the Manifold connector: look for maps_search_places, maps_get_place, maps_search_businesses, seo_get_serp and aeo_run_ai_answers (hosts often add a prefix such as mcp__manifold__). If the maps_* tools are missing while other Manifold tools are there, the Maps group is switched off on the app's Tools page; say so and use seo_get_serp for the local pack. If none are there, still answer from this skill with the by-hand methods in the references, and tell the user once how to connect: on the Manifold plugin's Connectors tab (Customize > Plugins in Claude), /mcp in Claude Code, or codex mcp login manifold in Codex; otherwise add the remote MCP server https://mcp.manifoldmcp.com/mcp and sign in. Other clients: https://www.manifoldmcp.com/docs/clients. Never invent a ranking, a review count or a measurement.

What the evidence says

The findings that change decisions. Sources and detail are in the references.

  1. Category and proximity decide the pack. The 2026 Local Search Ranking Factors survey (Whitespark, 47 experts scoring 187 factors, November 2025) puts primary category, proximity to the searcher, keywords in the business name, an address in the searched city, and being open at the time of search at the top. Google's own help page names relevance, distance and prominence and says no one can pay for a better local rank (ranking factors).
  2. The pack changes street to street. In our test three Columbus ZIP codes a few kilometres apart showed 8 different practices across their 9 pack slots for "dentist", and suburbs 8 to 30 km out shared none with the city's pack; on Maps a 3 by 3 grid 1.5 km apart put 16 different practices in its 27 top-three slots, and the downtown leader was top three in only 2 of the 9 cells (field notes). Maps and the search page's pack are not the same list either: for the same city they shared none of three. One rank check is one point; measure a grid (rank grid).
  3. The verified address anchors rankings; service areas do not move them. Hiding the address cost rankings and calls in Sterling Sky's test and restoring it brought them back a month later (run 2023, published September 2026); Whitespark saw a service-area business's rankings follow its address, not its service area (2024). Google caps service areas at 20, about two hours' drive, and banned countries and states as service areas in June 2025 (Business Profile).
  4. The Business Profile lost Q&A and gained a Gemini answer box. The Q&A API was retired on 3 November 2025 and the section came off profiles from December 2025; Ask Maps (Gemini, launched March 2026 in the US and India) answers questions about a place from its listing, reviews and website, and owners cannot write or correct its answers. The profile, the reviews and the website are now the only ways to shape them.
  5. Review rules tightened, and fake reviews carry fines. Since April 2026 Google's policy bars asking staff to hit review quotas or to get reviews that name a staff member; asking only happy customers (review gating) was already barred. The FTC's Consumer Review Rule (in force October 2024) bans fake reviews and incentives tied to sentiment at up to $53,088 per violation, with first warning letters to 10 companies in December 2025; the UK has banned fake and undisclosed incentivised reviews since April 2025. Google blocked or removed 292 million reviews and 13 million fake profiles in 2025 (reviews).
  6. Buyers want fresh, many and answered. Of 1,002 US adults (BrightLocal, February 2026), 74% read only reviews from the last three months, 47% will not use a business with fewer than 20 reviews, 31% want 4.5 stars or more (up from 17%), and 80% favour a business that replies to all reviews while 50% are put off by templated replies. Review count and recency help ranking; rating and replies mostly win the click.
  7. AI engines pick local businesses from listings, and pick few. Across 1.97 million citations for 1,355 locations, the Business Profile was 28.6% of all citations and Yelp 9.5%; Google's AI Overviews and AI Mode lean on the profile, ChatGPT on Yelp (BrightLocal, August 2026). In our runs Gemini cited only Google Maps listings (16 of 16). ChatGPT recommended 1.2% of 350,000 multi-location sites against a 35.9% local pack appearance rate (SOCi, January 2026) (AI answers).
  8. AI Overviews rarely show on "service in town" searches. They appeared on 15% of local-intent queries against 92% of informational ones, while the pack showed on 93% (Whitespark, 540 queries, May 2025); 21.8% against 75.8% in a 7,990-query study, where AI Mode answered every query and sent 79.8% of local citations to Google Maps listings (Steady Demand, vendor, August 2026). In our 12 local checks one overview appeared, on a cost question.
  9. Keyword-stuffed names still rank, and get profiles suspended. The survey ranks keywords in the business name third for the pack and keyword stuffing eighth among suspension risks. Name, category and storefront-to-service-area edits are the ones most likely to trigger re-verification (SOCi, March 2025), usually by live video. Report a competitor's stuffed name; do not copy it.

Does not hold up: "citations and NAP consistency are a top ranking factor" (their weight in the pack has fallen for years; they matter more for AI answers now); "add every town to the service area to rank there" (rankings follow the address); "rankings drop if reviews stop for three weeks" (traced to no study; the Sterling Sky case behind it is one 2022 client); "staff contests for reviews that name them are fine" (barred by Google since April 2026); "star rating drives pack rank" (count, recency and sentiment weigh more; rating mostly drives the click); "seed your own Q&A" (the section is gone); "one rank check from the office shows where we rank".

Unknown: how Ask Maps chooses which businesses to name; whether service areas have any ranking effect at all; whether the April 2026 suspension wave practitioners reported on service-area businesses was a policy change or a false-positive sweep (Google said nothing); how many calls AI answers send a local business.

References

Open the one the question needs.

The user asksRead
Why does that competitor outrank us on Maps? What actually moves the local pack?Ranking factors
Audit our Google Business Profile. Categories, hours, services, verification, a suspension, a service area businessBusiness Profile
Where do we rank across town? A grid or heatmap, tracking it weeklyRank grid
Who are our local competitors and how do we compare on reviews, categories and profile?Competitors
How do we get more reviews, reply to them, deal with fake ones, and stay legal?Reviews
Do ChatGPT, Gemini or Google's AI recommend us for local searches?AI answers
We are an agency: a list of local businesses to pitch (no website, few reviews, unclaimed)Prospect list
We want a 90-day local SEO planStrategy
What has Manifold seen in its own runs?Field notes

Using Manifold

  • Who Maps ranks at one point: maps_search_places (1 credit flat for up to 100 places, cached 24 hours) with keyword and either location (a city, read as one point at its centre) or latitude, longitude and zoom (14 a neighbourhood). One call per grid point; a 5 by 5 grid is 25 credits a keyword. Rows give rank (places only; ads take no number), rating, reviews, category, is_claimed and website. In our test 9 of 100 rows were other trades and 41 were unclaimed, most of them individual practitioners: filter before counting. status is open or closed at the moment of the call, by the listed hours, not whether the business still trades, and a row's category can be the one that matched the search rather than the primary (field notes).
  • The local pack on Google search: seo_get_serp (1 credit, 2 with ai_overview: true). Pack rows are type local_pack at ranks 1 to 3 and push organic result 1 to rank 4; they carry rating but no review count. Sponsored pack entries also come back as local_pack rows, with "Sponsored" in the snippet and often no url: drop them before counting. location takes a city or a ZIP ("43215,Ohio,United States"). The pack follows it; the organic rows of a bare search ("dentist") do not, at any level: at the United States, Columbus and Seattle no organic page was about the searched city, and four of nine were the same in all three. For local organic results put the town in the keyword ("dentist columbus ohio" returned nine Columbus-area pages). It is not Maps' top three: in Columbus the two shared none of three, in Austin two.
  • One profile: maps_get_place (2 credits, cached 7 days) by place_id, cid or name with the town. It returns categories, hours, rating_counts, review topics (at most 10), description, book_online_url and is_claimed. In 11 live profiles services and people_also_search were always empty and attributes held only accessibility items, even for the business whose services and related places fill the vendor's documented sample: the gap is upstream, so check those on the profile by hand and never report them as missing. Review text is not returned.
  • Everything in an area: maps_search_businesses (4 credits plus 12 per 100 rows returned, cached 7 days) filters a listings database by categories (Google ids in snake case), claimed, rating and review bounds. It is unranked and not live: updated_at ran 2 to 10 weeks old in our test, and claimed status and phone differed from live Maps on some rows, so confirm a prospect with maps_get_place before acting on it. max_reviews drops businesses with no reviews at all, and in healthcare most unclaimed rows are individual practitioners, not practices (prospect list).
  • AI answers: aeo_run_ai_answers, 2 credits a prompt each on chatgpt, gemini, ai_overview and ai_mode. Put the town in the prompt and keep location a country, except for "near me" prompts on gemini and ai_mode, which need the town or ZIP as location (with "United States" they were answered for Massachusetts and Colorado). A city location failed on ChatGPT with a misleading "rate limited" error, and Claude and Perplexity take a country only (AI answers).
  • Never post, reply, edit a profile, report a competitor or contact anyone on the user's behalf. Deliver the tables, the drafts and the steps. Never suggest buying, gating or writing reviews. Date every grid and store it in the host, since no monitor runs one; new reviews can be a reviews monitor.

Related skills

  • Organic rankings for the site, location pages and Search Console: seo. Crawl and indexing problems: technical-seo.
  • AI answers beyond local, measurement noise and wrong facts: ai-search. Local press, best-of lists and links: digital-pr and link-building.
  • Pitching, onboarding and reporting to local clients: agency. Finding the owner's email and the rules for contacting them: outbound.
  • What reviews say about customers' needs: voice-of-customer. The town's Facebook groups and the page: facebook. Local Services Ads and Maps ads: paid-ads. Chains and rivals beyond the pack: competitor-intel.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.