How ChatGPT decides which businesses to recommend

Two ways ChatGPT answers

When someone asks ChatGPT “Who’s the best roofer in Tampa?” or “Which accounting software should a small agency use?”, the answer comes from one of two places, and often both.

• What the model already knows. During training, the model read a huge amount of public text. Brands that were mentioned often and consistently across the web are “known” to it. This is why widely discussed brands can be named even when no search happens.
• A live web search. For questions that need current or local information, ChatGPT searches the web. Its help center explains that it rewrites the question into one or more targeted queries, sends them to its search providers, may run follow-up searches after reading the first results, and then writes an answer with citations.

For local and service questions, the search path almost always wins. That is good news: it means recommendations reflect what the web says about you today, and you can change that.

Where the information comes from

ChatGPT’s search draws on a mix of sources, and for local results it also uses third-party providers for business listings, maps and reservations. In practice, the pages that feed a recommendation usually include:

• Review and listing platforms such as Google, Yelp, Tripadvisor, Better Business Bureau and the directories that matter in your industry.
• “Best of” lists and local publications: city guides, trade publications, news coverage and comparison articles.
• Forums and communities where real customers discuss options.
• Your own website, especially pages that state plainly what you do, where, for whom and at what price.

Two technical details matter here. First, ChatGPT’s search features use a crawler called OAI-SearchBot. According to ChatGPT’s crawler documentation, sites that block it in robots.txt won’t be shown in ChatGPT search answers. Second, the help center points to Microsoft as one of its search providers, so being properly indexed in Bing (and listed in Bing Places for local businesses) is worth checking, not just Google.

What makes one business the pick

No one outside the company knows the exact weighting, and it changes as models are updated. But across many tests, the businesses that get named share a consistent profile:

1. Relevance and specificity. The business clearly matches the question: the right service, the right city, the right type of customer. Vague “we do everything” pages lose to pages that say exactly what is offered.
2. Consensus across sources. When several independent sites describe a business the same way (“family-owned, emergency service, specializes in flat roofs”), the engine can state that confidently. One source is a claim; five sources are a fact.
3. Reviews. Volume, recency, rating and, importantly, detail. Reviews that mention specific services and places give the engine language it can reuse.
4. Brand mentions and backlinks. Being named on lists, in local press, by partners and in industry directories signals that the business is real and notable. Links help those pages, and your own pages, get found.
5. Accessible, quotable facts. Hours, service areas, prices or price ranges, credentials and specialties written as plain text the engine can lift into an answer.

Why it names your competitors instead

If ChatGPT recommends a competitor and not you, the cause is usually one of these:

• They appear on more of the review sites, directories and “best of” lists that the search returns.
• Their reviews are more recent or more detailed, so they match the specifics of the question.
• Their website states the service and location explicitly, while yours implies it.
• Your business facts conflict across the web (old address, different phone numbers, a former name), which makes the engine less confident.
• Your site blocks or slows down ChatGPT’s crawler, or key information is locked in images, PDFs or scripts.

Each of these is fixable, and most can be checked in an afternoon.

How to become a recommendation

1. Check access. Make sure robots.txt, your firewall and your CDN allow OAI-SearchBot, and that your site is indexed in Bing as well as Google.
2. Write answer-ready pages. One clear page per core service (and per location if you serve several), opening with a direct description: what you do, where, for whom, how much, and why you are different.
3. Make your facts consistent. Name, address, phone, hours and service areas should match everywhere. Add Organization, LocalBusiness and Service structured data that matches the visible page.
4. Earn reviews on the right platforms. Ask every happy customer, on a schedule, on the review sites that show up when you search your own buyer questions.
5. Earn brand mentions and backlinks. Get listed in the directories and “best of” lists that already appear in answers for your category, and pursue local press, partners and associations.
6. Re-test and adjust. Ask your buyer questions regularly and record who gets named. Effort should follow the gaps.

How to check what ChatGPT says about you

Start with five to ten questions a real buyer would ask, phrased the way they would type them (“best family dentist in Plano that takes Delta Dental”). Ask each one several times, in a fresh chat with memory off, and note three things: whether you are named, who is named instead, and which sources are cited. Because answers vary from run to run, a single test proves little; patterns across runs are what count.

That is the same approach we use in our Ai visibility methodology, scaled up across every major engine and repeated monthly.

What you can’t control

Be wary of anyone who promises a guaranteed ChatGPT recommendation. Answers change with the wording of the question, the user’s location and chat history, and each model update. What you can control is the evidence: how clearly you describe yourself, how consistently the web agrees, and how many trusted sources vouch for you. Improve those, measure the results, and the odds move in your favor.