Local Business AI Visibility Checklist: Gemini, Apple Maps & Siri

Why One Checklist Is Not Enough

Local SEO checklists were built for one engine with one map. AI local search has at least four engines, and they do not share a map. Google's assistants read Google's places graph. Apple's read Apple's. Microsoft's read Bing's. ChatGPT reads whatever it can find, which in practice means the listings and review sites the other three already trust, plus your website.

That produces a specific failure: a business can be first in the Google local pack and absent from Siri, because Siri never looks at Google. Or present in ChatGPT because of a strong Yelp profile and absent from Gemini because the Google Business Profile has the wrong category. The fix for each gap is different, which is why the checklist is grouped by assistant rather than by tactic.

What Each Assistant Actually Reads

Read down the second column and the priority is obvious: three listings (Google, Apple, Bing) feed four of the five assistants directly, and Yelp feeds the review layer of three of them. Everything else on the list is supporting evidence for those. The ChatGPT local ranking guide covers the fourth row in depth; this checklist covers the rest.

Items 1-8: The Entity Foundation

Do these first. Every assistant checks them, and a failure here undermines everything below.

1. One canonical name. The exact legal or trading name, spelled and punctuated one way, with no keywords or city appended. "Evolve Plumbing" everywhere; never "Evolve Plumbing - Colorado Springs Drain Experts" on one listing.
2. One canonical address format. Suite numbers, abbreviations (St vs Street) and ZIP identical on every platform. Service-area businesses hide the address consistently everywhere, never on some listings and not others.
3. One phone number. The same primary number on every listing and on the site. Call-tracking numbers on directories break the match; if you must track, use one tracking number on the site only and keep listings on the main line.
4. One website URL. Same protocol, same www or non-www, same trailing slash, no UTM parameters on the listing URL. Point every listing at the homepage or the location page, and keep it the same one.
5. One primary category, chosen from the platform's own taxonomy. The most specific category that describes the core service, matched as closely as each platform allows. Category mismatches across platforms are the most common cause of an assistant recommending you for the wrong thing.
6. Hours that agree, including holidays. Assistants answer "open now" questions from listing hours. A Sunday mismatch between Google and Apple costs the Sunday customer.
7. A written 100-word description, reused verbatim. Plain language: what you do, for whom, where. The same text on Google, Apple, Bing, Yelp and the site's About page. Repetition is a feature; it is how assistants confirm they are looking at one entity.
8. A named owner or lead with a public profile. A real person, with the same name on the website, LinkedIn and any press. Entities with a visible human behind them get cited more; the E-E-A-T guide covers why.

Items 9-16: Google Business Profile For Gemini

Gemini, AI Mode and AI Overviews compose local answers from Google Business Profile before anything else. The Google Business Profile for AI search guide goes deep; these are the eight items that matter most.

9. Claimed, verified, and showing as the owner. An unverified profile is treated as unconfirmed data.
10. Primary category plus every relevant secondary category. Secondaries expand the questions you can be recommended for; irrelevant ones dilute confidence.
11. Services listed individually with descriptions and, where true, prices. Gemini answers "who does X near me" from the services list rather than from the description.
12. Attributes completed. Accessibility, payment types, amenities, "identifies as," service options. Attributes are structured data Gemini can filter on.
13. Twenty or more photos, real, recent, geotagged where possible: exterior, interior, team, work in progress, finished results. Assistants use photos to verify the business is operating and to match "what does it look like" questions.
14. Q&A seeded with the ten questions customers actually ask, answered by the owner. Google's AI reads Q&A as a source of facts about the business.
15. Weekly posts for eight weeks, then monthly. Recency is a confidence signal; a profile with no activity since 2024 reads as possibly closed.
16. Booking, quote or message link connected. Gemini increasingly completes the action; a profile with a booking link gets recommended over one that only has a phone number.

Items 17-24: Apple Business For Siri And Maps

Apple Business (the platform formerly called Apple Business Connect; one 2026 source reports the rename and merger on April 14, 2026) is the free tool at business.apple.com that creates your Apple Maps place card. Every Siri local answer, every Maps search, Spotlight, CarPlay and, as Apple Intelligence expands, its composed local answers, draw from that card. Apple does not host its own reviews; it shows ratings from Yelp, TripAdvisor and Foursquare, which is why items 21 and 38 exist.

17. Claimed and verified at business.apple.com. Unclaimed cards are built from third-party data feeds and are routinely wrong on hours and phone.
18. Category chosen from Apple's taxonomy, matched to the Google primary. Apple's categories differ from Google's; pick the closest, and record which you chose so the two do not drift.
19. Logo, cover image and 6-10 photos uploaded. Cards with owner-supplied images display ahead of cards using scraped ones.
20. Action links configured: order, reserve, book, menu, or quote, whichever applies. These are the buttons Siri can hand off to.
21. Yelp profile claimed and current, because that is where Apple pulls your rating from. A great Google rating does nothing for your Apple card.
22. A Showcase published and refreshed quarterly. Showcases are Apple's version of posts and the only recency signal on the card.
23. Holiday and special hours set before each holiday. Apple surfaces "closed today" prominently; a wrong closure on Apple Maps loses the customer standing outside.
24. UTM-tagged website link on the card, so Apple Maps referrals are separable in analytics. Apple's own insights are thin; this is how you measure the channel. One source reports a new "Suggested Places" surface arriving in iOS 26.5 in the US and Canada; if it ships, complete cards are what populate it.

The strategic point, which our Apple Intelligence guide makes for product brands and holds even more for local ones: Apple Maps has a fraction of Google's competitive density. Most competitors have never logged in. A complete card is cheap visibility in front of an iPhone-owning audience that skews older and higher-income.

Items 25-30: Bing, Yelp And The Rest

Copilot reads Bing Places and the Bing index. Alexa reads Yelp and Bing. ChatGPT reads all of it through search. These six items close the remaining gaps.

25. Bing Places claimed and verified at bing.com/forbusiness. Bing offers a one-click import from Google Business Profile; use it, then check the category mapped correctly.
26. Yelp profile complete: categories, hours, photos, the same 100-word description, and a response to every review. Yelp feeds Apple, Alexa and ChatGPT; it is the most under-maintained listing among businesses that do not think of themselves as "Yelp businesses."
27. TripAdvisor or Foursquare claimed where relevant. Hospitality, food and attractions need TripAdvisor; Foursquare data still feeds several aggregators.
28. Facebook page with matching NAP and hours. Not because assistants read Facebook much, but because data aggregators do, and aggregators feed the long tail of directories that assistants cross-check.
29. The five industry directories that matter for your category, with matching NAP. Angi and HomeAdvisor for trades, Healthgrades and Zocdoc for practices, Avvo for law, and so on. Assistants weight category-specific directories over generic ones.
30. A machine-readable summary of the business on your own site. A short llms.txt or an equivalent plain-text page that states name, address, phone, services, areas served and hours, so any crawler gets the facts without parsing a theme. Examples of what to include are coming in a later guide in this series.

Items 31-36: Your Own Website

Listings establish that you exist. The website establishes what you know, which is what assistants quote when the question is more specific than "near me."

31. LocalBusiness schema on the homepage or location page, using the most specific subtype (Plumber, Dentist, Restaurant), with name, address, telephone, geo coordinates, openingHoursSpecification, priceRange, areaServed, and sameAs links to every listing from items 9-29. The schema markup stack guide has the structure.
32. One page per core service, written in plain language, answering what it is, who it is for, what it costs or how pricing works, and how long it takes. Assistants cite pages that answer a question directly.
33. One page per primary service area if you serve more than one city, with something real on each (a project, a local detail) instead of a swapped city name.
34. An FAQ page with the questions customers ask before booking, marked up with FAQPage schema. This is the page assistants quote most for local businesses.
35. Crawlable by AI bots. robots.txt allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended and Bingbot; no CDN rule challenges them. A blocked crawler means a blank entity to that assistant. The AI crawler list covers the user agents.
36. Fast on mobile and readable without JavaScript. Most AI crawlers do not execute JavaScript. If your address and services only render client-side, the crawler sees nothing.

Items 37-40: Reviews, By Assistant

Reviews are the one signal every assistant weights heavily and each one reads from a different place. Volume matters less than recency, distribution and responses.

37. A review every week on Google, earned with a post-service link. Gemini reads velocity as a sign the business is active. Ten reviews a month for one month and none after reads worse than one a week.
38. Reviews on Yelp too, because Apple and Alexa read Yelp. Yelp filters solicited reviews aggressively, so the play is a Yelp link on receipts and signage rather than a request campaign.
39. Every review answered within a week, by name, in plain language. Assistants read responses as part of the entity's text. A response that restates the service and the area ("Glad the water heater install in Briargate went smoothly") is a fact the assistant can use.
40. Reviews that mention the specific service. Ask customers to say what was done. "Great service" tells an assistant nothing; "fixed our tankless water heater same day" is the sentence that gets you recommended for tankless water heater repair.

Test It: The 12-Query Audit

Before and after the checklist, run the same twelve questions on each assistant and record whether you appear, where, and what the assistant says about you. Three query types, four assistants.

The entity queries are the ones that tell you which source an assistant is using. When Siri says you close at 5 and you close at 6, your Apple card is wrong. When Gemini describes you as a general contractor and you are a plumber, your Google primary category is wrong. Fix the source, wait two to four weeks, re-run.

The 30-Day Order Of Operations

Most of the 40 items take under an hour each. The whole list is a working week spread over a month, plus the site pages, which take longer if they do not exist. A business with a decent site and a claimed Google profile is usually 60% done before starting; the remaining 40% is almost always Apple, Bing, Yelp and schema.

What Does Not Move The Needle

Since we sell this, here is what we would not charge you for.

• Keyword-stuffed business names. They violate every platform's guidelines, get suspended on Google, and lower entity confidence because the name no longer matches your license, your site or your invoices.
• Hundreds of directory citations. Past the three listings, Yelp, and your five category directories, additional citations are noise. Aggregator submissions that push the same data to 60 sites add little and create 60 places for it to drift.
• Review volume campaigns. A burst of 50 reviews looks synthetic to every platform and Yelp will filter most of them. One a week, mentioning the service, beats it.
• "AI optimization" tools that promise placement. No tool can place you in an assistant's answer. The tools that are worth paying for monitor whether you appear; the AI visibility tracking roundup covers them.
• A chatbot on your site. It does nothing for how assistants see you. Spend the money on the service pages instead.

Tracking What Changed

Three measurements, none of which need a tool.

The first measurement is the one that matters in the first 90 days, because listing changes take weeks to propagate and traffic is small. If the entity queries go from wrong to right on all four assistants, the checklist worked, and the referrals follow. The local AI referral data guide shows what those referrals look like at scale, and how small they are compared to the phone calls the listings themselves generate.

One last counterweight. For most local businesses, Google Business Profile still produces more calls than every AI assistant combined, and that will stay true for a while. The reason to do the other 32 items now is that they cost a week, they compound, and the businesses that do them in 2026 are the ones assistants will already be sure about when the volume shifts.