How to get cited by ChatGPT: a step-by-step tutorial

ChatGPT reads far more pages than it credits. In an Ahrefs analysis of 1.4 million ChatGPT prompts, only about half of the URLs ChatGPT retrieved while searching ended up cited in the answer (Ahrefs, April 2026). Learning how to get cited by ChatGPT is therefore two separate problems: getting into the pool of pages it retrieves, and then being the page it chooses to credit.

This tutorial covers both in seven steps. Each step says what to do, how to check it, and how strong the evidence is. You need access to your robots.txt file and CDN settings, an analytics tool, and about an hour for the first pass.

How does ChatGPT decide which pages to cite?

When ChatGPT searches, it turns the prompt into one or more narrower search queries, collects results, reads a subset of pages and cites the ones it relied on. OpenAI says ChatGPT search uses third-party search providers plus content from its partners (OpenAI), and its OAI-SearchBot crawler decides which sites are eligible at all. A citation needs eligibility, then retrieval, then relevance, in that order. The steps below follow the same order.

Step 1: Allow OAI-SearchBot in robots.txt

OpenAI runs separate crawlers for separate jobs. OAI-SearchBot controls whether your pages can appear in ChatGPT search answers. GPTBot collects training data. ChatGPT-User fetches a page when a person asks ChatGPT to open it. OpenAI says sites that opt out of OAI-SearchBot “will not be shown in ChatGPT search answers”, though they can still appear as plain navigational links (OpenAI crawler documentation).

Add or confirm these groups:

User-agent: OAI-SearchBot
Allow: /

User-agent: ChatGPT-User
Allow: /

Blocking GPTBot does not affect this. OpenAI says each setting is independent of the others, and that ChatGPT search takes about 24 hours to reflect a robots.txt change.

How to check: open yourdomain.com/robots.txt and look for any group that names OAI-SearchBot, or a User-agent: * group with Disallow: / that would apply to it. The robots.txt tester checks a URL against every AI crawler at once. For complete policies to copy, see robots.txt for AI bots.

Step 2: Make sure your firewall agrees with your robots.txt

A common failure: robots.txt allows AI crawlers, and the CDN blocks them anyway. Bot-management settings at the CDN, hosting firewall or a security plugin can challenge or reject requests that identify as AI crawlers before robots.txt is ever consulted.

How to check: the AI visibility checker requests your page with the OAI-SearchBot, GPTBot, ClaudeBot and PerplexityBot user agents and compares each response with a normal request. Treat a mismatch as a strong hint rather than proof, because real crawlers are verified by IP address, which no outside test can reproduce. Your server or CDN logs are the definitive record: filter for OAI-SearchBot and look at the status codes. OpenAI publishes IP ranges for each crawler if your firewall needs an allowlist.

Our view: this step finds more real problems than any content change, because a block here silently overrides everything else on this list.

Step 3: Get into the search pool

Being crawlable is not the same as being retrieved. In the Ahrefs dataset, pages that reached ChatGPT through its standard web search channel were cited at a rate of about 88%, compared with about 12% for news results and under 2% for Reddit threads. The authors put it plainly: if you want to be cited, “you need to be in that search selection pool.” Research shows this pattern; the study does not reveal which search provider supplied each result.

What gets a page into that pool:

1. The page is indexable. No noindex, a self-referencing canonical, a 200 status and no redirect chain.
2. The page is indexed by the major engines. Confirm with URL Inspection in Google Search Console and Bing Webmaster Tools.
3. The main text is in the HTML. Our view: server-render your content. Many AI crawlers fetch raw HTML and do not run JavaScript, so text injected client-side may never be seen.
4. The page ranks for the narrower queries ChatGPT is likely to issue. Conventional SEO is the entry ticket here, not a separate discipline.

Step 4: Match titles and URLs to the sub-questions

ChatGPT rarely searches for the prompt word for word. It issues narrower queries, a process usually called query fan-out. In the Ahrefs study, cited pages had titles closer in meaning to the user’s prompt than non-cited pages, and the closest match of all was between cited titles and the fan-out queries ChatGPT generated. Search results with readable, natural-language URL slugs were cited at a higher rate (about 90%) than those without (about 81%).

Research shows a correlation here, measured with open-source embedding models as a stand-in for ChatGPT’s own relevance scoring. It is still the most direct evidence available on what separates cited from uncited pages.

To apply it:

1. List the questions a buyer asks on the way to your answer. For “best payroll software for a small team”, that includes pricing, integrations, contractor support and setup time.
2. Check that each important page states its specific topic in the title, H1 and slug, in plain words. “Payroll software pricing for teams under 20” beats “Our pricing”.
3. Where a sub-question has no page, answer it inside the most relevant existing page before creating a new one.

The query fan-out explainer walks through a full worked example.

Step 5: Write passages that can be quoted

Once ChatGPT reads your page, it needs a passage that answers the sub-question cleanly. Under each question-led heading, give the answer in the first 40 to 60 words, then the detail. Use specific numbers with their sources. Name the people and documents you quote.

The KDD 2024 GEO paper tested this kind of editing. Adding statistics, quotations and cited sources raised a source’s visibility in generated answers by roughly 30–40% on the paper’s main metric, while keyword stuffing performed worse than doing nothing (Aggarwal et al., KDD 2024). Research shows, with a caveat: this was a benchmark built by the authors, not a test of ChatGPT, so read the effect size as an upper bound rather than a forecast.

Step 6: Get mentioned on the pages ChatGPT already cites

For most commercial prompts, ChatGPT leans on third-party sources: comparison articles, review sites, documentation and news. If your brand is missing from those, your own page has to win on its own.

1. Run your top prompts and note the sources ChatGPT cites. They are listed with each answer.
2. Group them: editorial lists, review platforms, forums, partner sites.
3. Work on being included through product data, original research, expert commentary or corrections of outdated information.

Research shows that brand mentions track AI visibility: in a study of 75,000 brands, branded web mentions had the strongest correlation (0.664) with appearing in Google AI Overviews (Ahrefs, May 2025). That study measured AI Overviews rather than ChatGPT, and correlation is not causation. Our view: earned mentions are worth having for their own sake and plausibly help across answer engines.

Step 7: Measure on a fixed prompt set

You cannot improve what you only check occasionally. Set up a repeatable baseline:

1. Write 20 to 30 prompts your buyers would plausibly type, from broad to specific.
2. Run each in ChatGPT with search on, in a clean session, and record whether your site is cited with a link, mentioned without a link, or absent. Note which sources were cited instead.
3. Repeat monthly. Answers vary between runs, so judge the trend across the whole set, not any single answer.
4. In analytics, filter for referrals from chatgpt.com. ChatGPT usually tags outbound links with utm_source=chatgpt.com, so these visits show up as their own source.

Prompt-tracking tools automate steps 2 and 3. Our comparison of AI visibility tools covers what each one measures.

What will not move ChatGPT citations

• An llms.txt file. OpenAI’s crawler documentation does not mention it. Our view: it is cheap to publish but has no documented role in ChatGPT search. The llms.txt guide covers the evidence.
• Schema added in the hope of AI citations. A matched test by Ahrefs in May 2026 found no measurable citation lift on ChatGPT after pages added JSON-LD. Keep structured data for rich results; do not expect it to earn citations. The ChatGPT SEO myths post has the details.
• Allowing GPTBot. Training access shapes what future models know, slowly and unmeasurably. It does not put your page into today’s search answers.
• Repeating the keyword. In the GEO benchmark, keyword stuffing lowered visibility.

What to do next

Do steps 1 and 2 today: they take minutes and the results are binary. Then choose five pages that should be cited for your most valuable prompts and work through steps 3 to 5 on those pages only. Record your step 7 baseline before you change anything, so you can tell whether the changes worked. For the same method applied to Perplexity, Claude and Google’s AI features, read the AI visibility guide.

Frequently asked questions

How long does it take to get cited by ChatGPT?

Crawler access changes are fast: OpenAI says robots.txt updates for OAI-SearchBot take about 24 hours to apply. Content changes take longer because the page has to be recrawled and then retrieved for the right queries. Plan on weeks, and measure against a baseline prompt set rather than single answers.

Do I need to allow GPTBot to be cited by ChatGPT?

No. OpenAI says GPTBot only controls whether content may be used for model training, and that each crawler setting is independent. OAI-SearchBot is the crawler that decides whether your pages can appear in ChatGPT search answers.

Why does ChatGPT mention my brand but not link to my site?

A mention can come from the model's training data or from another site it retrieved, such as a review or comparison page. A link to your site requires one of your pages to be retrieved during the search and selected as a source. Work on both: your own pages and the third-party pages ChatGPT already cites.

Does ranking in Google or Bing help with ChatGPT citations?

Usually, yes. OpenAI says ChatGPT search draws on third-party search providers, so pages that rank well in conventional search are more likely to be in the pool ChatGPT retrieves from. OpenAI does not publish the exact mix of providers.

How do I see ChatGPT traffic in Google Analytics?

Look for the referral source chatgpt.com. ChatGPT usually adds utm_source=chatgpt.com to the links people click, so these visits appear as their own source in GA4 and most other analytics tools.