How ChatGPT Picks Which Local Businesses to Recommend

A homeowner types “who's the best pet-safe pest control near me?” into ChatGPT. A few seconds later they have three company names, a sentence about each, and a suggestion to ask about warranties. They call the first one.

If you run the fourth-best pest control company in town, that call never happens, and you never find out it could have. So how does ChatGPT decide? Here is what's known, what's reasonable to infer, and what to ignore.

Two ways ChatGPT answers: memory and search

ChatGPT can answer from two places. The first is what the model learned during training, a large snapshot of the web from some point in the past. The second is a live web search, which ChatGPT runs when a question needs current, local information. OpenAI says its search feature draws on third-party search providers and content from partners.

Local questions (“near me”, “this week”, a city name) usually trigger a search. That's good news: it means what the web says about you today matters more than what it said years ago. It also means your business has to be findable by the search systems the AI relies on, not only by Google.

How a local answer gets built

Notice the last step. The reason ChatGPT gives for recommending you (“known for fast scheduling”) usually comes from somewhere. If your site and reviews never say what you're good at, the AI has nothing specific to repeat, and a competitor who spells it out gets the sentence instead.

What businesses that get named have in common

Run the same customer questions through ChatGPT, Gemini, Claude and Perplexity for a few weeks and look at who keeps getting named. Those businesses tend to share a few traits, and each one lines up with how the answer is built:

• They are specific. Their pages say exactly which services they offer, in which towns, for which kind of customer. “Residential window cleaning in Boise, Meridian and Eagle” beats “quality service you can trust”.
• Their facts match everywhere. The same business name, phone number, address or service area and services on their website, Google Business Profile, Yelp, Facebook and industry directories.
• Other people talk about them. Reviews, local news, community groups and “best of” lists give AI independent confirmation that the business exists and does good work.
• They answer questions out loud. Pricing ranges, how fast they can come, what's included, guarantees. AI quotes businesses that answer the question the customer asked.
• Their reviews say something. “Came out the next morning and found the leak” is evidence. “Great job!” isn't.

None of this is a trick. It's the same thing that makes a customer trust you: clear, consistent, specific information, confirmed by other people.

The research: how you write changes how often you're cited

The best public evidence comes from the 2023 GEO paper by researchers at Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi. They tested different ways of rewriting web pages and measured how often AI engines cited them. Adding citations, quotations from relevant sources and statistics worked best, raising visibility by up to 40% on their benchmark and up to 37% in a live test on Perplexity.

One finding matters a lot for small businesses: the sites that gained the most were the ones that weren't already at the top of Google. In one test, pages ranked fifth in search results saw the largest jump when they cited their sources. AI answers can give a smaller business a fairer shot than a page of links does, if its content is clear and backed up.

Is it different for Gemini, Claude and Perplexity?

Each assistant has its own search setup, so the same question can produce different names. Gemini and Google's AI Mode are built by Google and sit on top of Google's index and Business Profiles, so your Google profile carries extra weight there. Perplexity shows its sources next to every answer, which makes it a handy place to see which pages the AI is reading about your trade. Claude and ChatGPT use web search when a question needs current information.

That's why checking one tool once isn't enough. A business can be named by ChatGPT and missed by Gemini in the same week. The only reliable picture comes from asking all of them the same questions, regularly.

What to do with this

1. Check where you stand across the four main assistants. Here's a step-by-step method.
2. Write down what you want AI to say about you in one sentence, then make sure your homepage, service pages and Google profile say it plainly.
3. Fix inconsistencies. Old phone numbers, a former business name, or service areas that don't match confuse the cross-check.
4. Ask happy customers for specific reviews that mention the service and the town.
5. Measure again in a few weeks. Answers move, so track the same questions over time.

Want the full playbook? Read 10 GEO fixes for local service businesses, or start with the basics in What is GEO?.