How do I get my business cited by ChatGPT?

If buyers keep telling you “ChatGPT recommended” your competitor and nobody ever says it about you, you’re probably wondering how ChatGPT picks who to cite and whether there’s any way onto that list. I think there is, and it’s less mysterious than it looks, so here’s how ChatGPT chooses and the order I’d fix things in.

Where does ChatGPT get the businesses it names?

It pulls from two places, what it learned about brands back in training and what it finds in a live web search right before it answers, and the links you see come from that live search.

The most useful research I’ve found is a University of Toronto paper by Mahe Chen and Nick Koudas. They found every one of the 22,315 URLs ChatGPT cited had been returned by its own searches, so if your page doesn’t come back when ChatGPT searches, it can’t be cited no matter how good it is.

But the training side still counts for a lot. In the same paper, when a brand was already in ChatGPT’s prior knowledge it wrote that brand’s name into its searches 67% of the time, and only 11% of the time when it wasn’t. Sites the model already knew also had 1.60 times the odds of being cited at the same rank. So a competitor who’s been written about all over the web starts ahead, and the honest way to close that gap is to get written about too, which takes longer than anybody wants.

What searches does ChatGPT run when a buyer asks about my kind of business?

It chops the buyer’s question into a handful of shorter keyword searches, a lot of times with the year tacked on, so your pages need to answer those smaller searches and not only the big question.

The Toronto paper found ChatGPT splits a question into about 4 keyword-style searches and adds the current year to 23% of them. So a question about Dallas wealth managers for owners selling their company might become one search for wealth managers in Dallas, one about exit planning and one with 2026 on the end.

Ahrefs looked at 15 million of these fan-out searches and found 6.8% used the site: operator to search inside one website, and 75.9% of the most common site: searches pointed at a brand’s own site. After “2026,” the words ChatGPT added most were “official” and “price.” So sometimes ChatGPT goes looking for your official prices and facts on your own site, and if they aren’t there it quotes somebody else’s guess or skips you. Keep it in proportion though, cause Ahrefs also found only 14.4% of the 66.6 million free, logged-out responses it studied had fan-out searches at all.

Does ChatGPT use Google or Bing?

The studies don’t agree, so my take is be indexed and healthy in both, and set up Google Search Console and Bing Webmaster Tools before you touch anything else.

A 2025 Seer Interactive study of about 100 questions found more than 87% of ChatGPT search citations matched Bing’s top 20 results, against 56% for Google. That’s a small sample. The Toronto paper looked at ChatGPT’s own rewritten searches and found only 0.5% of the pages ChatGPT retrieved were in Bing’s top 10, and 0.7% for Google’s. They measured different things, and I’d rather tell you that straight than pretend one settles it.

Either way, a Google rank report won’t show where you stand in ChatGPT. Bing Webmaster Tools has a free AI Performance report showing which of your pages get cited in Microsoft’s AI answers and the grounding queries behind them, and while that isn’t ChatGPT data I’d still turn it on.

What does ChatGPT look for in the page it picks?

It leans toward pages that rank high in its own searches, come back across several of them, answer plainly near the top and were updated recently, and you can work on all of that from your own site.

Rank inside ChatGPT’s searches

The Toronto study found ChatGPT cites 43.8% of the pages it ranks first and only 3.5% of the pages ranked 11 to 20. A page that came back from several of its searches had 10.8 times the odds of being cited. And with ChatGPT citing only about 2 to 3 pages per answer the field is tiny, so I build pages that cover the buyer’s question plus the follow-ups, that way one page can show up across several searches.

An answer worth quoting

Most service pages open with a slogan, and a model can’t quote “trusted by families for decades.” It can quote your price range, your timeline, your licensing and your years in business if they’re written plainly and you’ve signed off on them. So I put a direct answer in the first few lines and the facts right after.

Recent updates

A 2026 Seer Interactive study that dated 7,683 pages found 73% of the pages ChatGPT cited had been updated within the past year. If your service pages haven’t been touched since the site launched, they’re starting behind, and the fix is real updates, not a new date at the top.

Words the bot can actually read

Vercel reported that none of the major AI crawlers currently render JavaScript, GPTBot included. So if your words only show up after a script runs, ChatGPT’s crawler may see a blank page, and the fix is to keep them in the raw HTML.

Google’s AI answers work a bit differently, and I get into those in how to rank in Google AI Overviews.

Do I need mentions on other websites to get cited by ChatGPT?

I’d say yes, since other sites shape what ChatGPT already knows about you and a lot of times they’re what it cites, but your own site still has to hold the facts.

Similarweb found Wikipedia made up 13.15% and Reddit 11.97% of U.S. ChatGPT citations in January and February 2026. Searchable, summarizing an Ahrefs study of about 75,000 companies, reported that branded web mentions correlated with AI visibility at 0.664, against 0.218 for backlinks. Those are correlations and not proof, but they line up with the Toronto finding that known brands get searched for by name.

For most established companies I’d start by fixing the listings that already exist. Directories and review profiles with an old phone number or a service you quit offering years ago are all feeding the same model.

What should I skip?

I’d skip the shortcuts like llms.txt files, special AI markup and keyword stuffing, and I haven’t seen a reliable effect from any of them in the data I’ve read.

SE Ranking looked at nearly 300,000 domains and found no correlation between having an llms.txt file and how often a domain got cited by AI tools. A Metavert digest of AI search research cites a July 2026 survey of 45 GEO studies that concluded no reviewed technique shows a stable, cross-platform causal effect. So anybody selling one trick for ChatGPT is out ahead of the evidence. And if your pages read like a chatbot wrote them, I get into that in how to make content not sound like AI.

What’s the step-by-step plan to get cited by ChatGPT?

Measure first, fix the pages and then what other sites say about you, and measure again so you know if anything moved. Here’s the order I’d go in.

1. Write down 20 to 50 questions your buyers actually ask before they hire someone like you, in their words, since that’s what ChatGPT builds its searches from.
2. Run each one through ChatGPT three times on different days and write down who gets named, so you’ve got a real starting point.
3. Set up Google Search Console and Bing Webmaster Tools and submit your sitemap to both, and turn on Bing’s AI Performance report while you’re there.
4. Rewrite the page that should answer each question so the answer sits in the first few lines, in plain words, with your real numbers on it.
5. Add the follow-up questions a buyer would ask next to that same page, so it can show up across more of ChatGPT’s searches.
6. Put the words in the HTML itself and not behind tabs, sliders or scripts, cause the AI crawlers don’t run JavaScript.
7. Fix your directory, review and industry listings until every one says the same true thing.
8. Update the pages that matter at least once a year with real changes, not just a new date.
9. Rerun the same questions every month and compare against where you started.

I’d run each question more than once cause the answers move. The Metavert digest cites an April 2026 study that found roughly 65% of the sources an engine cites for a query change from one day to the next. That’s annoying when you’re measuring, but it also means the list isn’t locked.

Where should I spend effort first?

What does it cost to wait?

Waiting won’t sink a healthy company next month. What it costs is position, slowly. ChatGPT only cites about 2 to 3 pages per answer, so every question a competitor owns today is one less seat for you. And since known brands get searched for by name far more often, the companies getting talked about now get easier to find later, while pages nobody updates slowly age out.

The common outcome is a firm that still does fine on Google and gets left off the AI shortlist without anybody noticing, and losing every answer in your market is the rare worst case. The next step is cheap though. Ask ChatGPT five of your buyers’ questions today and write down who it names.

How I know this can work

I’ve done this on a business I own. My own ecommerce brand’s Bing AI citations went from 9 a month in April 2026 to 1,275 a month in August 2026, measured in the Bing Webmaster Tools AI performance report. That report covers Microsoft’s AI answers and not ChatGPT directly, and the brand also runs paid ads, so I don’t claim all of that growth came from the pages. When I checked rankings live in July 2026 it ranked #1 on Google for the “best” search for its product, with 18 of 22 target terms in the top 3. You can see how each of my earlier results was measured. Hailder House is new and has no finished client builds yet, so these numbers come from my own work, they aren’t typical, and yours will be different.

For how ChatGPT fits with Claude, Perplexity, Gemini and Google’s AI, read my full guide to AI visibility, and for why I think your Google rankings still matter underneath it all, see whether SEO is dead because of AI.