Local SEO in ChatGPT: How AI Search Recommends Nearby Businesses

The stakes are rising fast. In BrightLocal's March 2026 report, 31% of US consumers said they had used ChatGPT for a business recommendation in the past year, and 23% had used Google's AI Mode. Yet the AI shortlist is tiny. SOCi's 2026 Local Visibility Index, which covered more than 350,000 locations of 2,751 multi-location brands, found that ChatGPT recommended 1.2% of them. The same locations showed up in Google's local 3-Pack 35.9% of the time.

Most local SEO advice still stops at the Google Business Profile. This guide covers the sources each AI assistant uses for "near me" answers as of September 2026, graded by how sure anyone can be. Then it gives a three-part framework, a prompt test and a worked example. For the full business fact sheet and listing audit, see our guide to optimizing your business for AI search engines. For one assistant only, read how to get cited by ChatGPT as a local business. For daily tasks, see eight ways AI helps small businesses with local SEO.

How AI Assistants Answer a "Near Me" Question

Classic local SEO is about ranking pins on a map. An AI assistant does something else. It writes a short answer, and to do that it runs a chain of steps. We grade every step in this guide by its evidence:

• Documented: the assistant's maker, or its data partner, says so on its own pages or in a filing.
• Observed: an independent test saw it. That shows a pattern, not how the system is built.
• Inferred: a reasonable guess that nobody has confirmed.

Here is the chain for ChatGPT, which is the assistant most people ask about.

1. It works out where you are. ChatGPT uses a rough location from your IP address. Since 26 March 2026, people can also share their device location, which is off by default. Saved memories can shape the search too. Documented.
2. It rewrites your question. OpenAI's search help page gives the example: "What are some good restaurants near me?" asked from San Francisco becomes "top restaurants San Francisco," which goes to a search partner. This is a form of query fan-out. Documented.
3. It looks up places. The same page says ChatGPT may share your location with "trusted third-party providers that help provide more local information, such as nearby business listings." It doesn't name them. In August 2026, Suganthan Mohanadasan logged a separate places lookup in ChatGPT's search calls that takes a location and returns business names. Documented that providers exist; observed how the lookup runs.
4. It reads ratings and reviews. Yelp told investors that its ratings and reviews now power ChatGPT's local answers "in relevant categories." Documented by Yelp.
5. It checks the web and writes the answer. In the same logs, ChatGPT checked each business name with follow-up searches, then drew a map with pins. Your website only joins this step if OpenAI's crawler, OAI-SearchBot, can reach it. Observed; the crawler rule is documented.

Why the AI shortlist is so short

A map shows twenty pins. A chat answer names a handful of businesses. That's the gap behind SOCi's 1.2% figure. SOCi also found that in retail, only 45% of the brands most visible in classic local search were among those AI platforms recommended most.

So near-me visibility in AI is its own local SEO job. It isn't a side effect of ranking in the map pack, and a strong Google profile alone won't cover it.

Which Map, Listing and Review Sources Each Assistant Uses

Every assistant draws on a different map. The table shows what each maker or data partner says, checked on 29 September 2026. It's the heart of local SEO for AI search, because a listing only helps with the assistants that read it.

Siri: Apple Maps and Apple Business

Apple's iPhone guide tells people to ask Siri something like "Find coffee near me" to search Apple Maps. You manage your Apple Maps place card in Apple Business, which replaced Apple Business Connect on 14 April 2026 and moved existing data over on its own. Apple says the new platform helps businesses reach customers "across Apple Maps, Mail, Wallet, Siri, and more."

Apple doesn't build its business data alone. Its Maps acknowledgements page credits business listings data from Foursquare, Tripadvisor, Yext, Localeze, Uberall and others, plus "Reviews from Yelp." So a wrong fact on your place card can come from one of those feeds, not only from you. For wider web questions, Siri AI (in beta in English since 14 September 2026) can draw on the web and on Applebot's crawl. Our voice search optimization checklist covers the Applebot rules, and our guide to how Apple Intelligence routes Siri queries covers when Siri hands a question to ChatGPT.

Copilot: Bing Places

Microsoft is direct. When it launched AI citation reports in Bing Webmaster Tools, it said accurate business information matters "when AI experiences surface answers to location-based queries," and told businesses to register with Bing Places so their details stay "eligible for inclusion in AI-generated responses." Our Copilot guide covers the rest of Bing's setup.

Google: one profile for three AI surfaces

Google's AI optimization guide says Business Profiles "can help your products and services to be visible in both AI responses and other Google Search results." Gemini uses public Google Maps data for addresses, ratings and hours, and Ask Maps draws on reviews from more than 500 million contributors. Most businesses have this pin already.

Perplexity: partner deals, no public list

Perplexity publishes no local source list on its own pages. Its partners do the talking: Yelp data came in March 2024, and Tripadvisor announced a partnership in January 2025 covering more than 11 million listings. How much either feeds answers today is unknown, so for local SEO in Perplexity your Yelp and Tripadvisor pages are the parts you can shape.

ChatGPT: what the evidence supports

A common line in local SEO guides is that ChatGPT pulls local data from Apple Maps, Bing Places and Yelp. The evidence splits three ways.

Yelp: supported. Yelp's quarterly filing for the period to 30 June 2026 says its "ratings and reviews recently began powering ChatGPT's local experience in relevant categories." Its August blog post adds reviews, photos and business details shown in the chat. Then an independent test backed it up. Ben Fisher of Steady Demand ran 2,880 prompts across 12 US metros and 12 business types in one 72-hour window in August 2026. Yelp sat in the data behind ChatGPT's business cards in 95.83% of runs.

Apple Maps: not supported. OpenAI never mentions it. In the same test, Apple Maps supplied 0.00% of business-card data and had a 0.14% share of visible citations. Apple Maps matters a lot for Siri. For ChatGPT, nothing public says it matters at all.

Bing Places: unproven. Microsoft is a named ChatGPT search partner, and Bing Places feeds Bing. But OpenAI doesn't say that listing data flows through, and Bing had a small share in the test. Claim Bing Places because it's free and feeds Copilot, not because it's proven for ChatGPT.

What about Google? OpenAI doesn't say ChatGPT uses Google data. In the test, Google supplied 4.62% of business-card data. Fisher warns that "citation share is not data influence," so a source can shape answers without showing up. Treat every share above as a snapshot of one month.

The Three-Pin Framework for Local SEO in AI Search

Most businesses have pin zero of local SEO: a Google Business Profile. The Three-Pin Framework adds the three pieces that reach assistants Google doesn't feed. It covers Apple, Bing and your own markup, grades each one, then adds the one piece the evidence says ChatGPT needs.

Pin 1: Sync your Apple Business place card

1. Sign in at business.apple.com and claim your location. If you used Business Connect, check that your data moved over.
2. Copy your name, address, phone, hours and category from your Google profile. Copy and paste; don't retype.
3. Add photos and a custom action, such as book, order or reserve.
4. Add a showcase if you run a current offer.
5. Check it: search your name in Apple Maps, then ask Siri for your category near your street.

If a fact on the card is wrong and you didn't enter it, check Yelp and the listing feeds Apple credits. Fix the fact at the source, or it may come back.

Pin 2: Sync your Bing Places listing

1. Go to bing.com/forbusiness. Microsoft moved Bing Places there in October 2025.
2. Import from your Google profile. Microsoft says the new import is better at "preserving key attributes like business name, hours, and contact details."
3. Read every imported field anyway. Imports carry over old errors as faithfully as good data.
4. Verify the listing, then ask Copilot for your category in your town.

Pin 3: Add LocalBusiness JSON-LD with geo coordinates

Be clear about what this pin does. Google says its AI features need no special markup. No assistant says it reads coordinates from your web page for a near-me answer. So treat the markup as cheap insurance: it helps Google read your details, and it gives every crawler one clean copy of the facts on your listings.

Google's LocalBusiness docs set the rules. Use the most specific type. name and address are required. For geo, "the precision must be at least 5 decimal places." Five decimals of latitude is about 1.1 meters, since one degree of latitude spans roughly 111 kilometers. That's enough to put the pin on your door, not the block.

Here's the core for the made-up bike shop in our worked example below:

{
  "@context": "https://schema.org",
  "@type": "BicycleStore",
  "name": "Quillfen Cycle Repair",
  "url": "https://quillfencycles.example",
  "telephone": "+1-509-555-0147",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "240 Larch Row",
    "addressLocality": "Spokane",
    "addressRegion": "WA",
    "postalCode": "99202",
    "addressCountry": "US"
  },
  "geo": {
    "@type": "GeoCoordinates",
    "latitude": 47.66132,
    "longitude": -117.40415
  },
  "openingHoursSpecification": [
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": ["Tuesday", "Wednesday", "Thursday", "Friday"],
      "opens": "10:00",
      "closes": "18:00"
    },
    {
      "@type": "OpeningHoursSpecification",
      "dayOfWeek": "Saturday",
      "opens": "09:00",
      "closes": "16:00"
    }
  ]
}

To get your coordinates, drop a pin on your front door in any map app and copy the numbers it shows. Our free schema generator builds the block from a form, and Google's Rich Results Test checks it.

The ChatGPT add-on: a claimed Yelp page

The evidence points to one more listing for ChatGPT. Claim your Yelp page, then fix the category, hours and phone. Yelp says its content also reaches Apple Maps and Microsoft Bing, so one fix lands in three assistants' data. Yelp's rules differ from Google's: it asks businesses not to request reviews at all. Our business listings guide covers each platform's review rules.

Reviews, Ratings and Local SEO in AI Answers

Stars work like a gate in AI answers. Locations ChatGPT recommended in SOCi's study averaged 4.3 stars. That is a correlation across big chains, not a documented cutoff, and small businesses with fewer reviews may see a different pattern. Customers check, too: in BrightLocal's report, 88% said they fact-check reviews that AI tools cite.

Each assistant reads a different pile of reviews:

• ChatGPT: Yelp ratings and reviews, per Yelp's filing.
• Gemini and Ask Maps: Google ratings and reviews from Maps.
• Siri and Apple Maps: Apple's own ratings, where people tap like or dislike for things like service and quality. The feature isn't in every country. Apple also shows reviews from Yelp.
• Perplexity: Yelp and Tripadvisor, per the partner deals.

For local SEO in AI search, track three numbers on each platform every month: your rating, your review count and the date of your newest review. A 4.9 on Google means little to ChatGPT if your Yelp page shows 3.6 from five old reviews.

Apple Maps also asks recent visitors to rate a place on its own, using on-device suggestions. That means your Apple rating grows from real visits, which is one more reason to get the place card right first. How to ask for reviews within each platform's rules is covered in our business listings guide.

The Near-Me Prompt Test

You can't see inside an assistant, but you can see its answers. That makes testing the most honest part of local SEO for AI. The Near-Me Prompt Test is a fixed set of six prompts that shows which assistants name you and which facts they get wrong. Run it before you change anything, then again four to six weeks after.

How to run it

1. Pick four assistants your customers use. For most US businesses: ChatGPT, Gemini or AI Mode, Copilot, and Siri with Apple Maps. Add Perplexity if your trade is on Yelp or Tripadvisor.
2. Use a logged-out window, or a ChatGPT temporary chat set to Unpersonalized, so saved memories don't steer the answer.
3. Write down the location setting: IP only, or device location shared. OpenAI says a VPN can move your apparent location.
4. Run each prompt three times per assistant. Answers change between runs.
5. For each answer, record whether you're named, your position, each fact shown, and any source label, such as a Yelp logo.

Score two numbers per assistant. Mention rate is the answers that name you divided by the runs of prompts 1–5. Fact accuracy is the share of prompt-6 answers that state the fact correctly, with a missing fact counted as a miss. For panel sizes and noise, see our guide on how to measure GEO. Our free AI Visibility Prompt Kit drafts extra prompts with a scorecard.

Worked Example: Quillfen Cycle Repair

Quillfen Cycle Repair is a made-up, two-person bike repair shop in Spokane, Washington (quillfencycles.example). On Google, its local SEO looks strong. Its profile is complete, with 4.8 stars from 212 reviews. It never claimed Apple Business or Bing Places. A customer created its Yelp page years ago, and it shows nine reviews and last year's Saturday hours. The website has no LocalBusiness markup. Every number below is illustrative.

The owner runs the Near-Me Prompt Test. Prompts 1–5 in four assistants, three runs each, make 5 × 4 × 3 = 60 answers. Prompt 6 adds 4 × 3 = 12 more.

On prompt 6, 7 of 12 answers gave the right Saturday closing time, a fact accuracy of 58%. The four wrong answers from ChatGPT and Copilot each named the old Saturday closing time. The one from Siri gave no hours at all.

The pattern matches the source table, which is the point of doing local SEO per assistant. The assistant fed by the complete Google profile names Quillfen most. The three fed by listings nobody claimed name it least.

The fix order

That's 20 + 40 + 15 + 30 = 105 minutes, or under two hours of work. Yelp goes first because it's the only fix that reaches more than one assistant. The markup goes last because nobody documents its effect on AI answers.

The owner re-runs the same 72 prompts five weeks later, with the same settings. Local SEO changes take time to reach listing feeds, so an earlier re-test can mislead. The goal isn't a promised score. It's to see whether the three low assistants now name Quillfen, and whether the old Saturday hours are gone.

Where Rankbox Fits in Local AI Search

Rankbox doesn't do listing-side local SEO. It doesn't claim listings, manage reviews or edit your Apple, Bing or Yelp pages, and it doesn't track AI citations today. The Three-Pin work and the prompt test stay with you. The free Prompt Kit covers the manual test.

What Rankbox does is the website half of local SEO for AI. Answer-Space Research finds the questions customers ask ChatGPT, Perplexity and Google about your trade, with volumes as model estimates. Rankbox then writes sourced articles in your voice that answer them, like "how much does a bike tune-up cost," and they reach your site through Rankbox's API. The Business plan is $49.50 a month with a 7-day free trial. See how local businesses use Rankbox or plans and pricing.