See what makes AI recommend your business.

How AI Decides
What Businesses to Recommend

By Lesli Rose · Updated October 30, 2026 · 9 min read

When someone asks ChatGPT "who's the best plumber near me?", the answer isn't random. AI assistants look for the same handful of things every time: clear information about your business, and other people vouching for you. If they find both, you get named. If they don't, a competitor does.

I've run more than 50 visibility reports for coaches, software companies, vet clinics, breeders and local service businesses. In each one I ask ChatGPT, Perplexity, Claude and Google the questions a real customer would ask, and I note who gets named and which sources the answer leans on. The same patterns show up again and again. This article walks through them.

Where AI Gets Its Information

An AI assistant has two places to look.

The first is its training data, a huge snapshot of the web taken before the model was released. Think of it as long-term memory. If your business was well known across the web before that snapshot, the model may already know who you are.

The second is live search. When you ask a question, most assistants now search the web, read the top results and build an answer from them. This is where a review from last week or a page you published last month can make a difference.

Both reward the same thing: a business that shows up the same way, in several trusted places, over time. Training data reflects your past. Live search reflects your present. You want to be strong in both.

The Five Things AI Looks At

These work together, so don't read the order as a ranking. Miss one and the others carry less weight.

1. Structured Data (Schema Markup)

Schema is a small block of code that tells machines, in plain fields, what your business is, where it is, what you offer and when you're open. Without it, AI has to work it out from your page text. With it, there's nothing to guess. It's one of the first things I check on any site.

2. What Other People Say About You

This carries the most weight. In an AirOps analysis of 21,311 brand mentions, about 85% of AI citations came from third-party sources, not the brand's own website. Think Google and Yelp reviews, "best of" articles written by someone else, Reddit threads, directories and local news. One source is an opinion. Several independent sources saying the same thing is something AI is comfortable repeating.

3. Clear, Useful Pages

Does your site answer the questions customers actually ask? Who you serve, what you do, how it works, what it costs, where you are. A page full of slogans gives AI nothing to quote. A page that answers a real question in plain words gives it a sentence it can lift straight into an answer.

4. The Same Details Everywhere

AI tries to work out that the business on Google, the one on Yelp and the one on your website are all the same business. If your name, address or phone number differs from one place to the next, that gets harder, and your signals get split. Matching details everywhere is dull work, and it pays off.

5. Signs You're Still Active

When an assistant searches live, recent activity helps. A business whose last review was two years ago looks quieter than one getting reviews every month. Freshness won't replace the other four, but it makes all of them count for more.

How the Main AI Tools Differ

They all care about the same five things, but each one finds and shows information a little differently.

ChatGPT

ChatGPT runs its own web searches and doesn't rely only on Google, so being easy to find in other search engines, including Bing, widens your net. In my reports, the sources it leans on most for business questions are review sites, roundup articles and Reddit. It has also started adding warnings when a "best of" list was written by one of the companies on it, which is one more reason not to rank yourself.

Perplexity

Perplexity searches the web for every question and shows its sources right under the answer. That makes it the easiest place to see why you were or weren't named. If you only test one tool, start here and read the citations.

Google AI Overviews

AI Overviews draw on Google's own search index, so your regular SEO work feeds them directly. Your Google Business Profile, your page structure and your schema all play a part. If you already rank well on Google, you have a head start here.

Claude

Claude answers from its training data and can search the web when that feature is on. When it searches, it cites the pages it used, so plain, factual pages with your details stated clearly give it something solid to quote.

What Doesn't Get You Recommended

Knowing what doesn't count can save you money:

  • ›Ad spend. Paying for ads doesn't buy you a place in the answer itself. Some AI tools now show ads beside their answers, but those are labelled and kept separate from who gets recommended.
  • ›Follower counts. A big Instagram following doesn't make AI more likely to name you. What people write about you matters more than how many people follow you.
  • ›An old domain on its own. Having owned your web address for 15 years doesn't help by itself. What's on the site, and what others say about you, is what counts.
  • ›Looks alone. AI reads your words, not your design. A plain site with clear answers beats a beautiful site with little to say. Design still matters for the people who visit, so don't neglect it; just don't expect it to earn AI mentions.

Why It Builds on Itself

Every new review, mention, listing and useful page adds to the last. When AI finds you described the same way in five places, it's fairly confident. When it finds you in fifteen places, with recent activity, it's confident enough to name you without hedging.

That's why starting sooner helps. A competitor who starts six months after you isn't just six months behind on time. They're six months behind on reviews, mentions and pages that all support each other. I saw the same thing with Google in the early 2000s: the businesses that did the groundwork early held their spots for years.

What to Do Next

Your customers are already asking AI who to call. The good news is that the rules are knowable. Clear structured data, real reviews and mentions, pages that answer questions, matching details everywhere and steady activity are all within your control.

Start with one test. Ask ChatGPT and Perplexity the question your best customer would ask, using your town and your service. See who gets named, then look at which sources the answer cites. That list is your to-do list.

Frequently Asked Questions

Does AI use the same ranking factors as Google?

Not exactly. There's a lot of overlap, and a strong Google presence helps. But AI assistants lean much harder on what other sources say about you: reviews, roundup articles, directories and press. Good SEO is the starting point, not the whole job.

Can I game AI recommendations?

Not for long. Google and the AI companies filter fake reviews, stuffed pages and paid placements, and ChatGPT has started warning users when a ranking was written by the company being ranked. Real reviews, real press and honest pages are what last.

How often does AI update its knowledge?

A model's built-in knowledge stops at its training cutoff, usually several months before release. But the main assistants can all search the web live, so a review from last week can show up in an answer today.

What's the most important factor for AI visibility?

What other people say about you. In the AirOps analysis, about 85% of AI citations came from third-party sources. Your own website still matters, because AI goes there to confirm details, but the recommendation itself usually comes from somewhere else.

Find Out If AI Recommends You

I'll ask ChatGPT, Perplexity and Google AI the questions your customers ask, then show you who gets named, which sources they trust, and what's missing for you.

Run Your Visibility Report