How AI assistants decide which local businesses to recommend
ChatGPT recommends far fewer local businesses than Google's map pack. What 2026 data says about who gets picked, where information comes from and what to fix.
Ask Google for a plumber and the map pack shows three businesses, with a long list below. Ask ChatGPT and you may get two or three names, and nothing else. The shortlist is far shorter, which means the gap between being recommended and being invisible is far wider.
The clearest data on this so far comes from SOCi’s 2026 Local Visibility Index, which analyzed more than 350,000 locations across 2,751 multi-location brands. It found that “ChatGPT recommends just 1.2% of brand locations”, against “an average 35.9% appearance rate in Google’s local 3-Pack” (SOCi, January 2026).
That study covers chains, not independent businesses, and one study is not a law of nature. But the direction is consistent with what we see when we test local prompts: AI assistants are much more selective than a results page.
Where the assistants get their local information
Each assistant draws on different sources, and only some of them are documented publicly.
| Assistant | What is publicly confirmed |
|---|---|
| Google AI Overviews and AI Mode | Google tells businesses to keep “Business Profile information current” for these features (Google) |
| ChatGPT | Yelp licenses “reviews, photos and other business information” to OpenAI for ChatGPT answers, announced July 2026 (Axios). ChatGPT also searches the web when it answers. |
| Perplexity | Searches and cites the web, with sources shown. |
You will read confident claims elsewhere about exactly which map provider or index powers each assistant. Many contradict each other, and few cite the platforms themselves. We would rather tell you what is confirmed and work from that.
The practical conclusion holds either way: your Google Business Profile and your Yelp listing are now both direct inputs to AI answers, alongside your website and what others write about you.
What gets a business picked
Based on SOCi’s findings and the way these systems are built, four things do most of the work.
1. Reputation, as a gate. The locations ChatGPT recommended in SOCi’s data averaged a 4.3-star rating, and SOCi describes consumer sentiment as “one of the strongest gating factors for AI visibility”. An assistant putting its name to a recommendation behaves like a cautious friend: it avoids the 3.4-star option even if it is closer.
2. Consistent, complete details. SOCi again: “Inconsistent or incomplete listings reduce AI confidence and can remove brands from consideration entirely.” Different phone numbers, old hours or a missing category on one platform weaken every platform.
3. Specific, quotable information. “Emergency plumber, Omaha, typically on-site within 90 minutes, upfront flat-rate pricing” can be lifted straight into an answer. “We pride ourselves on quality” cannot. The more precisely your site and listings say what you do, where and for whom, the more often you match a specific question.
4. Mentions beyond your own site. Local news, trade directories, community discussion and independent “best of” lists give an assistant corroboration. A business only described by itself is a weak recommendation.
What to fix first
| Order | Fix | Why it comes first |
|---|---|---|
| 1 | Claim and complete Google Business Profile and Yelp | Both are confirmed inputs to AI answers |
| 2 | Make name, phone, address or area and hours identical everywhere | Inconsistency can remove you from consideration |
| 3 | Build a steady review habit with real customers | Ratings act as a gate |
| 4 | Give each main service its own specific page | Specific questions need specific answers |
| 5 | Earn a few local mentions: press, associations, directories | Corroboration from outside your own site |
Our Google Business Profile checklist covers step 1 field by field, and our guide to getting recommended by ChatGPT covers access and measurement.
How to check where you stand
Ask the assistants the way your customers do, and name the place, because results depend on location:
- “Who is the best [service] in [town]?”
- “I need a [service] in [neighborhood] this week. Who should I call?”
- “Is [your business] in [town] any good?”
Run each question several times on ChatGPT, Gemini, Perplexity and Google, note who is named and how you are described, and repeat monthly. Answers vary between runs, so one screenshot proves nothing in either direction. Our guide to measuring AI visibility explains how to make this systematic.
Questions people ask
Why does Google show my business but ChatGPT doesn’t? Because the shortlist is so much shorter. In SOCi’s data, Google’s 3-pack showed locations about 30 times as often as ChatGPT recommended them. Being “good enough” for a map listing is not the same as being one of the two or three a model will name.
Do I need a Yelp profile now? If you serve local customers, yes. Yelp’s licensing deal with OpenAI makes its reviews and business details a confirmed source for ChatGPT’s local answers.
Can a small independent business compete with chains in AI answers? Yes. Assistants weigh reputation and specificity, not size. A well-reviewed independent business with complete, consistent details and specific service pages is often a stronger answer than a chain location with a thin profile.
Does location sharing change the results? Results depend on where the person asking is, or the place they name. That is why testing should always include the town or neighborhood.