How ChatGPT Recommends Local Businesses

Nobody outside OpenAI knows the exact recipe, and OpenAI does not publish one. What is documented is the mechanism: when ChatGPT searches the web it retrieves results and writes an answer with links, so it tends to name businesses whose facts, reviews and mentions it can find and that agree with each other.

This article explains why ChatGPT answers the way it does, as part of the complete AEO guide. If you first want to see what it says about your business today, check what ChatGPT says is the two-minute version.

What is documented, and what is not?

OpenAI's help center says ChatGPT can decide to search the web, turn a question into one or more searches, and use the results to write an answer that links to its sources. The same page says ChatGPT may share your location with trusted third-party providers that supply local information, such as nearby business listings. It also warns that search results and citations can be incomplete, outdated or incorrect.

OpenAI's crawler documentation adds one fact you can act on. A crawler named OAI-SearchBot is used to surface websites in ChatGPT's search features, and a site that disallows it will not be shown in ChatGPT search answers. If your website blocks that crawler, nothing else in this article matters yet.

What OpenAI does not publish is the weighting: how much reviews count against mentions, which directories it prefers, how many names it lists. Any article that hands you a precise formula is guessing. The rest of this one reasons from how retrieval-based answers work and labels its inferences as inferences.

What happens when a customer asks?

Here is the sequence a retrieval-based answer plausibly follows. The first four steps match how OpenAI describes search. The last one is inference, because nobody outside the company can watch it.

• A customer types a full sentence with a place and conditions: "Which tire shop in Weslaco can do two tires today without an appointment?"
• ChatGPT decides whether it needs current information. A question about one town and today's availability depends on facts that change, so searching is a likely move.
• It turns the question into one or more searches and retrieves results: web pages, listings and review text.
• It reads what came back and writes an answer, often with links to the sources it used.
• Which businesses make it into the answer depends on how well their information fit the question and agreed across sources. This step is inference.

Which signals plausibly feed a local recommendation?

These come from reasoning about how retrieval works, not from a published ranking formula. They are the things an assistant can find and compare, which makes them the things worth fixing.

• Business facts that agree everywhere. Name, address, phone, hours and categories on your website, your Google Business Profile and the directories that list you. If your site says open Saturday and a directory says closed, an assistant reading both has to pick one or skip you.
• Reviews and what they say. Google's Business Profile help says more reviews and positive ratings can help local ranking. OpenAI has not published how ChatGPT weighs reviews, but review text is public, readable and full of specifics.
• Mentions on other sites. A chamber of commerce page, a supplier's dealer list or a local news story that names you shows that someone besides you says you exist.
• Clarity of your own pages. A page for each service that states the city, the hours and the conditions a customer cares about, such as walk-ins, insurance or Spanish-speaking staff.
• Readability. Text in the page rather than locked in an image or a PDF, and no block on OAI-SearchBot.

Say a taquería in Edinburg lists Monday to Saturday hours on its website but shows a year-old Sunday schedule on its Facebook page. A customer asks which taquería in Edinburg is open Sunday morning. The assistant finds two versions of the truth. It may cite the wrong one, hedge, or name a competitor whose information is consistent. Nothing is wrong with the tacos. The record is muddy.

Mentions follow the same logic. Backlinks vs brand mentions goes into what counts when another site names you, with or without a link.

What can you do with this?

Read your own footprint the way an assistant would. Search your business name plus your city, open the first ten results, and write down what each says about your hours, address, services and phone. Every mismatch is a fix. Correct the source first (your profile and your website), then the directories.

Reviews are the signal you can grow steadily. How to get more Google reviews covers asking in a way Google allows.

Assistants also differ from one another. Perplexity, Claude and Gemini each have their own way of finding sources, so a change that shows up in one may not show up in another. That is why tracking several assistants over several weeks beats one check. The Growrank dashboard does that weekly across ChatGPT, Gemini, Claude and Perplexity; see plans and the free trial. If the vocabulary in this space confuses you, AEO vs SEO vs GEO sorts it out.

Frequently asked questions