How to get your business cited by ChatGPT: a practical checklist

The short answer

ChatGPT is now a place buyers ask before they search. OpenAI says it has 900 million weekly active users, as reported by TechCrunch in February 2026. That figure is company-reported and unaudited, and it counts weekly users of every kind, not buyers in your category. It still means the assistant's answer to "who should I talk to about this" is a new shortlist you are either on or off.

This is the practical companion to our guide to generative engine optimisation. It covers the crawler rules, the index question, the content traits, the off-site work, and a checklist you can run this week.

How does ChatGPT decide which websites to cite?

When ChatGPT searches the web, it sends one or more queries to a search index, reads the returned pages, and cites the passages it used. It favours pages it can crawl, that answer the specific sub-question directly, that carry concrete facts, and that belong to brands other sites already talk about.

Three findings shape the checklist.

First, ranking helps but is not the whole story. In Semrush's study, almost 90% of ChatGPT-cited pages ranked at position 21 or lower for related queries (Semrush, June 2025, vendor study). Ahrefs found only 8.0% of ChatGPT citations were in Google's top 10 for the prompt, across 15,000 long-tail queries (Ahrefs, August 2025).

Second, ChatGPT is biased towards fresher pages. Ahrefs' study of about 17 million citations found ChatGPT the most freshness-biased of seven platforms, citing URLs 458 days newer on average than organic results (Ahrefs, July 2025).

Third, reputation beats link counts. Across 75,000 brands, YouTube mentions correlated with ChatGPT visibility at 0.737 and branded web mentions at 0.664, against 0.266 for Domain Rating (Ahrefs, December 2025). The same study found ChatGPT the most accessible engine for smaller brands.

Which OpenAI crawlers do you need to allow?

Allow OAI-SearchBot if you want to appear in ChatGPT search answers. GPTBot is a separate crawler used only for model training, and blocking it has no effect on search. ChatGPT-User fetches pages when a person asks ChatGPT to open a link. Each is controlled separately in robots.txt, so you can allow search and refuse training.

OpenAI describes OAI-SearchBot as "used to surface websites in search results in ChatGPT's search features" (OpenAI). Its documentation says a site that disallows it will not appear in ChatGPT search answers, although navigational links may still show.

Two practical notes. OpenAI says robots.txt changes take about 24 hours to reach its search systems (same source). And robots.txt is only half the gate. From 15 September 2026, Cloudflare blocks training and agent bots by default on ad-monetised pages, while leaving search bots allowed (Cloudflare). A firewall or CDN rule can block OAI-SearchBot even when robots.txt allows it, so check both.

A worked example: the wildcard trap

This is the pattern we look for when PostCliq reads a robots.txt. Suppose a site has these three groups:

The owner meant to refuse training. Because the wildcard group was never cleaned up after launch, the site also refused ChatGPT search, Perplexity and every other bot without a named group. Under the robots.txt standard, a crawler follows the most specific group that names it and otherwise the wildcard. PostCliq's audit evaluates robots.txt separately for each named AI agent, including GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended, and shows the verdict for each. It reports a block as a policy choice rather than a defect, because many blocks are deliberate. What it flags is the gap between what you meant and what the file says.

Does ChatGPT use Bing or Google results?

ChatGPT search relies on its own crawler, OAI-SearchBot, and the studies disagree on how closely its citations follow Bing. For short head queries the overlap with Bing is high. For long-tail and follow-up queries it is low. Keep Bing indexing healthy, because it is cheap, but do not assume Bing rankings alone decide citations.

The two best-known studies point in different directions. Seer Interactive found 87% of SearchGPT citations matched Bing's top results, against 56% for Google, in a small study of 100 queries (Seer Interactive, February 2025). Ahrefs, looking at 15,000 long-tail queries, found only 8.1% of ChatGPT citations in Bing's top 10 (Ahrefs, August 2025).

Both can be true. Head queries follow the index closely. Fan-out queries, the follow-up searches ChatGPT runs for you, reach deeper and wider. The cheap wins are the same either way: verify your site in Bing Webmaster Tools, submit your sitemap, and ping IndexNow when you publish or update. Bing's new AI Performance report, in public preview since 10 February 2026, shows how often your pages are cited in Microsoft Copilot and Bing's AI summaries. It does not cover ChatGPT, but it is the closest free proxy you have.

The checklist: twelve checks to get cited by ChatGPT

Work through these in order. The first four are access, and none of the content work matters if the crawler cannot reach you. Most checks take minutes. The off-site work takes a quarter.

What kind of content does ChatGPT quote?

ChatGPT quotes self-contained passages that answer one question directly, early on the page, with concrete facts. Definitions phrased as "X is", short answers under question headings, comparison tables and sourced statistics are easy to lift. Long introductions, vague claims and key facts hidden in images or tabs are not.

The evidence here is mostly vendor research, so weigh it accordingly. Surfer found the top three ChatGPT factors in 2026 all concern the first paragraph (Surfer, correlational). The Princeton GEO study, a controlled lab benchmark, found adding quotations lifted visibility from a baseline score of 19.5 to 27.8 and adding statistics to 25.9, while keyword stuffing fell to 17.8 (Aggarwal et al., KDD 2024). That is roughly a third more visibility for statistics (25.9 divided by 19.5 is 1.33), on a simulated engine rather than ChatGPT itself.

Microsoft's guidance for AI answers makes the same point in plain terms: descriptive question-style headings, one-to-two-sentence answers, lists and tables, and no important text inside images (Microsoft Advertising, October 2025).

For a service business, the pages that most often deserve this treatment are the ones buyers ask about: what you do, who it is for, what it costs, how long it takes, and how you compare. If your pricing page says "contact us" and nothing else, an assistant has nothing to quote, and it will quote a competitor's page that gives a range.

Why off-site mentions decide whether ChatGPT names you

ChatGPT repeats what the web says about you. If only your own site describes your business, you are one voice. If reviews, forums, partners, directories and videos describe you consistently, you become a safe answer. For most businesses this is the slowest part of getting cited and the part with the most leverage.

Where ChatGPT looks has also moved. In Profound's 2025 data, Wikipedia was ChatGPT's most cited source at 7.8% of citations (Profound, vendor study). Semrush then recorded ChatGPT's Reddit citations falling from about 60% of responses to about 10% in mid-September 2025, while LinkedIn rose steadily on every platform (Semrush, November 2025). The lesson is not "post on Reddit". It is that any single channel can be turned down overnight, so spread your presence.

The Similarweb figure is the caution. Visits to generative AI platforms grew 70% year on year to 9.5 billion a month, but ChatGPT's share of those visits fell from about 76% to about 53% between June 2025 and May 2026 (Similarweb, vendor panel data). Optimising for ChatGPT alone is optimising for half the market. Everything on this checklist also helps with Perplexity, Copilot and Google's AI features, which is the argument for doing it as normal good practice rather than as a ChatGPT project.

How do you know if it is working?

Track three things monthly: whether ChatGPT names you for a fixed set of buyer questions, how many sessions arrive from chatgpt.com in GA4, and whether those sessions become enquiries. Expect small numbers. Judge the trend over a quarter, not a week, and keep the test identical each time.

AI referral volume is still small for most sites, and it converts well. At Ahrefs, AI search sent 0.5% of visits but 12.1% of sign-ups in a 30-day window in June 2025 (Ahrefs). That is one brand-aware software company, so do not borrow its multiple. The broader Ruler Analytics dataset, over 110 million sessions in mostly UK lead-generation businesses, found AI referral traffic converting at 5.8% against 4.9% for organic search (Ruler Analytics, 2026).

The monthly prompt test is set out step by step in how to check your AI visibility, and the questions and scoring sheet are in the AI visibility check: ten buyer questions. If you also want to understand the Google side, what AI Overviews do to your search traffic covers what can and cannot be measured there.

What to do next

1. Read your robots.txt today. Confirm OAI-SearchBot and ChatGPT-User are not blocked by a named group or a leftover wildcard.
2. Check your CDN. Open your Cloudflare or firewall bot settings and confirm AI search bots are allowed.
3. Verify Bing Webmaster Tools and submit your sitemap. Turn on IndexNow if your platform supports it.
4. Rewrite the opening of your five most important pages so the first sentence names the service and the first paragraph answers the buyer's main question.
5. Replace three vague claims per page with a number and a linked source.
6. Plan one off-site mention a month: a partner page, a podcast, a review, a LinkedIn post with substance.
7. Set a baseline with the ten buyer questions check, then repeat it in 30 days.
8. Run a free PostCliq audit to see your AI crawler policy per bot, whether your text is readable without JavaScript, and what stops visitors becoming enquiries. The scope is on the AI search visibility page.

Frequently asked questions

Can I pay to be cited by ChatGPT?

No. OpenAI offers no paid route to organic citations in ChatGPT search answers. OpenAI does operate a separate OAI-AdsBot to check pages submitted as ChatGPT ads, but that is advertising, not a citation. Citations come from crawlable, indexed pages that answer the question well and from your reputation across other sites.

Does blocking GPTBot remove my site from ChatGPT?

No. GPTBot collects content that may be used to train OpenAI's models. OAI-SearchBot is the crawler behind ChatGPT search, and OpenAI controls the two separately. You can block GPTBot to opt out of training and still allow OAI-SearchBot so your pages can appear, with citations, in ChatGPT search answers. Check that no wildcard rule blocks both.

How long does it take to appear in ChatGPT?

OpenAI says robots.txt changes take about 24 hours to reach its search systems. New or rewritten pages depend on crawling and indexing, which can take days to weeks. Off-site reputation builds over months. Run the same buyer questions monthly and look for a trend across a quarter rather than expecting a change within a week.

Is ChatGPT SEO different from Google SEO?

Mostly no. The same foundations apply: crawlable pages, indexing, clear answers, sourced facts and a trusted brand. The differences are emphasis. ChatGPT cites many pages outside Google's top 10, favours fresher content, and leans heavily on what other sites say about you. So ChatGPT SEO puts more weight on opening paragraphs, freshness and off-site mentions.

Does ChatGPT show where its answers come from?

When ChatGPT searches the web it usually shows source links alongside or within the answer, and those are the citations this checklist aims at. Answers written from the model's training data, without a web search, often show no sources. That is why being mentioned widely on other sites matters too, because it shapes what the model already knows about you.

How do I track ChatGPT traffic in Google Analytics 4?

Look in GA4 traffic acquisition for sessions with chatgpt.com as the source. Treat the count as a floor, not a total, because some assistant visits arrive without a referrer and are recorded as direct. Pair the session count with your enquiry or lead key event, so you see whether those visits convert, not only whether they arrive.