How ChatGPT, Gemini and Google's AI choose local businesses
Ask ChatGPT, Gemini or Google's AI Mode for a good plumber nearby and you get a few names back, not a long list. Which names, and why, depends on the assistant. Studies published in 2026 found that each one leans on different sources, that they rarely agree with each other, and that the same question asked twice can get a different answer.
They recommend far fewer businesses than Google's local pack shows
SOCi's 2026 Local Visibility Index covered more than 350,000 locations belonging to more than 2,700 brands with 50 or more locations each. As Search Engine Land reported it, 35.9% of those locations appeared in Google's local 3-pack, while the three AI assistants tested, ChatGPT, Gemini and Perplexity, recommended between 1% and 11% of them: 7.4% on Perplexity and 1.2% on ChatGPT. That gap is where the report's headline came from: being recommended by ChatGPT was about 30 times harder than ranking in the local pack.
Search Engine Land named one reason: the assistants often had a business's facts wrong. Profile accuracy averaged 68% on ChatGPT and Perplexity, against 100% on Gemini, which pulls directly from Google Maps data.
Each assistant reads different sources
BrightLocal's August 2026 study ran five prompts from nine points around each of 1,355 business locations in the US, the UK and Australia, 60,970 checks in all, and recorded nearly two million citations. Where each assistant got its local facts:
| Assistant | Its most cited sources |
|---|---|
| Google AI Overviews | Google Business Profiles first, then Yelp, Facebook and Tripadvisor |
| Google AI Mode | Google Business Profiles first, then Yelp, Facebook and Tripadvisor, from the widest range of sources of the three |
| ChatGPT | Yelp, which appeared in 80% of its answers, then Tripadvisor, OpenTable and the Better Business Bureau |
A separate study by Steady Demand, reported by Search Engine Land in August 2026, looked at 14,472 citations from 1,487 local searches across 50 US metro areas. Gemini mostly cited businesses' own websites, while ChatGPT relied more on Reddit and business directories. Run on the same searches, the two cited mostly the same sites only 8% of the time and named the same top business just 4.2% of the time.
Google is bringing the same kind of answer into Maps itself. On March 12, 2026 it announced Ask Maps, which answers questions about places in conversation, and a Search Engine Land test that April found it moving from plain listings toward recommendations.
The same question gets different answers
Steady Demand also found that identical searches often produced different recommendations. Answers change from one asking to the next and from one assistant to another, so one screenshot of one answer says very little. What means something is a pattern: how often a business is named when the same questions are asked again and again, over weeks.
What this means for a local business
- Your Google Business Profile is a main source for Google's own AIs. AI Overviews and AI Mode both cited it more than any other source, and Gemini, which pulls from Google Maps data, had every business's facts right in SOCi's report. Keep the profile complete and correct.
- ChatGPT reads directories and forums. Yelp above all in BrightLocal's study, Reddit and directories in Steady Demand's. A listing that is missing or wrong there is what ChatGPT has to work with.
- Check each assistant on its own. In Steady Demand's study ChatGPT and Gemini named the same top business only 4.2% of the time, so being named by one says little about the others.
- Ask often and judge the trend. Ask the same questions on a schedule and count how often the business is named, rather than reacting to a single answer.
None of these studies shows that a particular change will get a business recommended. They show where the assistants looked for their facts and how rarely they agreed.
The short version
AI assistants recommend far fewer local businesses than Google's local pack shows, each one draws on different sources, and their answers change from one asking to the next. Track each assistant separately, ask on a schedule, and judge how often you are named over time.
Sources
- Danny Goodwin, Search Engine Land, AI local visibility is up to 30x harder than ranking in Google: Report, January 28, 2026.
- Near Media, Local AI study, AI Mode onramp, G-queries down, squandered tech trust, January 2026 (the study's size).
- Search Engine Land, Calls and clicks keep falling as Google Maps becomes the destination, August 31, 2026 (the shares and the accuracy figures).
- Mike Hawkes, BrightLocal, Top sources and directories for local AI search, August 26, 2026.
- Danny Goodwin, Search Engine Land, Business websites dominate Gemini citations in local AI search: Study, August 19, 2026.
- Search Engine Land, Google Maps turns exploration into a conversation with Ask Maps, March 12, 2026.
- Search Engine Land, Google Ask Maps is moving from listings to recommendations, April 14, 2026.