RESOURCES // LOCAL SEARCH

How do ChatGPT, Claude and Perplexity pick which local businesses to mention?

By Ricky Ybarra · Updated 2026-10-10 · 4 min read

The short answerWhen someone asks an AI assistant for a local business, it usually runs a web search behind the scenes, reads a handful of the pages that come back, and writes a summary from them. So it can only mention businesses whose details are published in clear text on pages it can reach: your own site, directories, review sites and local articles. There is no paid placement to buy and no way to guarantee a mention.

What happens when somebody asks

An assistant has two sources of knowledge. One is what it learned in training, which is general and out of date for local detail. The other is live search: it sends a query to a search engine, opens some of the results, and reads them the way a person would, only faster.

For a question like "who repairs garage doors near me on a Sunday", the live search does most of the work. The assistant is looking for pages that say, in words, that a business repairs garage doors, covers that area and is open on Sunday. Then it picks a few to name and often links to where it read each fact.

Where the information comes from

The assistant reads whatever the search returns. For local questions that tends to be a mix.

  • Your own website, if it can be found and read.
  • Business directories and map listings.
  • Review sites and trade directories.
  • Local news, community pages and "best of" lists.
  • Forum and social posts where people recommend a business by name.

What makes a business easy to mention

Nobody outside these companies knows exactly how each assistant weighs sources, and the companies change their systems often. What can be observed is simpler: an assistant can only repeat what it was able to read, and it prefers facts it finds stated the same way in more than one place.

  • Services named specifically, in text. "Tankless water heater installation", not "all your plumbing needs".
  • Towns and areas written out, not only shown on a map image.
  • Hours, including emergencies and weekends.
  • Prices or price ranges where you are willing to publish them.
  • The same name, address and phone number everywhere you are listed.
  • Recent reviews on public sites that mention the kind of work you did.

What gets in the way

Some sites are close to invisible to an assistant even though they look fine to a person. Text baked into images cannot be read reliably. A site that shows nothing until a lot of script has run may come back empty to a simple reader. A site that blocks AI crawlers in its robots.txt file, which is a small text file of instructions for automated visitors, has asked not to be read.

Vague writing is the other barrier. "Quality service at fair prices since day one" tells an assistant nothing it can use to answer a question.

How to see where you stand

Ask the assistants yourself. Use the questions a customer would use, with your trade and your town in them, in ChatGPT, Claude, Perplexity and Google's AI answers. Note who is named and, where links are shown, which pages the assistant read.

Do this a few times. Answers vary from one ask to the next, and they change with the wording. You are looking for patterns: which sources keep being cited, and what those pages say that yours does not.

What you cannot control

You cannot pay to be included in the written answer, you cannot submit your business to an assistant, and anybody who says they can guarantee you a mention is guessing. What you can do is make your facts clear, consistent and readable, which also helps with ordinary search and with human visitors.

That is the approach I take at Novyx. Sites I build state services, towns and hours in plain text with structured data behind them.

Questions people ask

Can I pay ChatGPT to recommend my business?

Not for a place in the written answer. Assistants build answers from what they find and read. Any advertising these products offer is separate from that and is labeled as advertising.

Do Google reviews affect AI answers?

Reviews on public pages are one of the things an assistant may read, and assistants built on a search engine inherit some of that engine's local results. Exactly how much weight they carry is not published.

Why does the AI recommend a competitor and not me?

Usually because it found their details stated clearly on pages it read, and did not find yours. Look at the sources it cites and compare what those pages say with what your pages say.

Is this different from SEO?

Mostly it is the same groundwork: clear pages, consistent listings, real reviews. The difference is that an assistant quotes specific facts, so stating them plainly matters even more.

Keep reading