How to Get Cited by ChatGPT

That is the whole job. Here is each part.

First, drop the idea of a ranking

There is no leaderboard inside ChatGPT. No position three. Nobody can sell you one, and if someone offers to guarantee you a spot, they are selling something that cannot exist.

In February, Marcus Sheridan met two partners in a remodeling business who were thrilled that ChatGPT had named them among the best remodelers in their city. He ran the same search on his own phone. They were not there. Same question, same week, different person, different answer.

ChatGPT answers are shaped by who is asking, what the model has seen before, and whether it decided to search the web at all that time. So stop asking "where do I rank" and start asking "how often does my name come up, across lots of people and lots of phrasings."

Step one: check ChatGPT is allowed to read your site

Most small businesses skip this step, and it is the only one that can take you to zero on its own.

OpenAI runs several crawlers with different jobs. GPTBot collects content for training models. ChatGPT-User visits pages when a person asks ChatGPT to go look at something. The one that matters for citations is OAI-SearchBot. In OpenAI's own words, sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers. That is as close to a rule as anyone gets. You can read the crawler list at OpenAI's bot documentation.

This got timely this week. On 15 September 2026, Cloudflare added a setting called Disallow AI Training, meant to let a site refuse training while staying in search. Its press release names Apple, Google and Microsoft as accountable crawlers that stay allowed, and says other mixed-use crawlers are blocked if a site owner blocks training. OpenAI is not named on that list. Sites with the old blanket block were moved over automatically.

So if your site sits behind Cloudflare and someone once ticked the box to block AI bots, go look today. Open your robots.txt in a browser. If OAI-SearchBot is disallowed, that is your answer, and no amount of content work will fix it. OpenAI says robots.txt changes take about a day to take effect.

Step two: write the answer, not the brochure

Once ChatGPT can read you, it has to find something worth lifting.

Ahrefs looked at the top 1,000 pages ChatGPT cited in September 2025 using its Brand Radar data. Wikipedia took 29.7 percent. Homepages were 23.8 percent. Only about a third of cited pages fell into the kinds of page you can actually create, like news, reviews and blog posts. Most citations go to places you do not control. Plan around that.

The last third is real, though, and the pattern in it is simple. ChatGPT lifts a passage it can quote without stitching four paragraphs together. So on every page that matters, put one clear question in the heading and answer it in two or three sentences underneath, in words a stranger would understand. Include the boring specifics: the city, the price range, the response time, the brands you service, the year you started. Those are what make a passage worth quoting.

One page, one question. A single page trying to cover fifteen topics gives ChatGPT nothing clean to pull.

Step three: get named where ChatGPT already looks

If you are a local business, this is where most of your leverage sits, and it moved this summer.

On 23 July 2026, OpenAI signed a licensing deal with Yelp. ChatGPT gets Yelp reviews, ratings, photos and business information, plus Yelp's Request a Quote flow for local services. Yelp's CEO said Yelp branding and links appear when OpenAI uses its content. Yelp is also named in OpenAI's own help pages, alongside OpenTable and Resy, as a reservation partner.

Before that deal, two local SEO vendors published on where ChatGPT got its local facts. Local Falcon said Bing Places was one of the biggest sources and that there is no direct Google Business Profile integration. Whitespark, across 153 queries in 17 categories and nine US cities, found Facebook and Yelp were the most common review sources in Bing Places. Both are vendor studies from 2025 and the Yelp deal probably shifted the picture, so treat them as a starting map, not a current one.

The practical version: your Yelp page, Bing Places listing and Facebook page are not leftovers from 2014. Claim them, fill them in properly, and make the wording match your site. If you are an MSP, add the industry directories your buyers read. Then get your name onto the third-party pages that already answer your buyer's question, the local roundups and the trade write-ups.

Step four: if you sell products, the feed is the door

Ecommerce works differently enough to matter.

Profound, a vendor that tracks AI answers, reported this month that ChatGPT Shopping swung hard toward merchant product feeds. Across roughly 1.76 million tracked prompt runs in July, feed sources went from 8.26 percent to 61.54 percent in a few days in mid July, and sat near 65 percent by 3 September. Over the same window the unique merchants appearing fell from 13,524 to 10,607. Fewer stores, showing more often. It is vendor data from a company selling tracking, and nobody has replicated it.

OpenAI's position is worth reading closely. It says product results are selected independently by ChatGPT and are not ads, and are not influenced by OpenAI partnerships. It also says merchants can apply for direct feed access, and that Shopify stores are already integrated through Shopify Catalog. So a feed gets your accurate product data in front of the system. It does not buy you the answer.

If you are on Shopify, your job is data quality: titles, prices, availability, descriptions. If you are not, look at whether a direct feed is worth applying for.

Step five: do not build on one source

In August, Reddit's share of ChatGPT citations fell from 3.83 percent to 0.52 percent in about a week, an 86 percent drop, according to Promptwatch.

Nobody agrees on why. The popular explanation is an 8 August change where ChatGPT began using site-specific searches far more heavily. But the big drop came on 14 August, six days later, which that does not cover. Promptwatch says it cannot rule out a problem in its own data collection. Something similar happened in September 2025, and Kevin Indig put that one down to a Google change, not to Reddit at all.

You cannot control that. What you can do is spread the surface. Your own site, your Yelp and Bing listings, the directories in your trade, a few third-party pages that mention you by name. If one of them goes quiet next month, you are still there.

What to do this month

Open your robots.txt and confirm OAI-SearchBot is allowed. Check your Cloudflare settings if you use it. Pick your ten most valuable buyer questions and give each one its own page with a plain two-sentence answer near the top. Claim and fill Yelp, Bing Places and Facebook. Then ask ChatGPT those ten questions yourself, from a logged-out browser, once a week, and write down what it says. That log is your measurement. There is no dashboard that beats it.

If you sell B2B software rather than services or products, the buying committee and review-site dynamics are different enough to need their own playbook, so read how B2B SaaS gets cited by ChatGPT instead. If you would rather have someone run the check for you, that is what our audit does.

Sources

• OpenAI, Overview of OpenAI Crawlers, developers.openai.com/api/docs/bots
• OpenAI Help Center, ChatGPT Search
• OpenAI Help Center, Shopping with ChatGPT Search
• Cloudflare, Have it both ways: stay discoverable in search while disallowing AI training, 15 September 2026
• Cloudflare, Cloudflare Helps End the Search-or-AI-Training Tradeoff, press release, 15 September 2026
• Search Engine Land, ChatGPT gains access to Yelp reviews, ratings and photos, 23 July 2026
• Axios, Yelp partners with OpenAI to surface reviews in ChatGPT, 23 July 2026
• Axios, Reddit fades from ChatGPT citations, 20 August 2026
• Search Engine Journal, Why Reddit's ChatGPT Citation Drop Isn't Fully Explained, 19 August 2026
• Search Engine Journal, ChatGPT Shopping Results Lean Hard On Product Feeds, 9 September 2026, reporting Profound data (vendor)
• Profound, How ChatGPT sources the web, 3 February 2026 (vendor)
• Ahrefs, ChatGPT's most cited pages, Brand Radar analysis of the top 1,000 pages cited in September 2025
• Whitespark, Want to rank in ChatGPT? Focus on these review sites, 15 September 2025 (vendor)
• Local Falcon, ChatGPT local search data sources, 5 November 2025 (vendor)
• 21 Hats Morning Report, We're getting recommended on ChatGPT, 5 February 2026