How to get cited in ChatGPT
ChatGPT reaches hundreds of millions of weekly users and increasingly answers buying-intent questions directly. Getting your brand named or cited inside those answers is a distinct discipline from Google SEO. This guide covers how ChatGPT sources information and the concrete steps to become one of the sources it trusts.
How ChatGPT decides what to cite
ChatGPT pulls from two places: its trained model (shaped by how widely and consistently your brand appears across the web) and live web results via ChatGPT search, which is retrieved by OpenAI's own crawlers and historically leans on the Bing index. Optimizing for it means being both broadly-referenced enough to be 'known' and crawlable enough to be retrieved in real time.
Crawler access (OAI-SearchBot & GPTBot)
ChatGPT search retrieves live pages with OAI-SearchBot; GPTBot informs the trained model. If your robots.txt blocks them, you are invisible to ChatGPT regardless of content quality.
Brand entity strength
The model recommends brands it has seen described consistently across many independent sources — reviews, directories, press, forums. Breadth and consistency of mentions matter more than any single page.
The Bing index
ChatGPT search has historically relied on Bing for web results, so being indexed and ranking in Bing is a practical prerequisite for real-time citation.
Answer-shaped, factual content
ChatGPT quotes concise, unambiguous statements — clear definitions, specs, and direct answers — far more readily than marketing prose.
How to optimize for ChatGPT, step by step
- 1
Allow OpenAI crawlers in robots.txt
Confirm OAI-SearchBot, GPTBot, and ChatGPT-User are not blocked. Verify with the Robots.txt AI-Crawler Checker — a single accidental Disallow removes you from ChatGPT.
- 2
Get indexed in Bing
Submit your sitemap in Bing Webmaster Tools and fix Bing indexing issues. ChatGPT search leans on Bing, so Bing visibility is a practical requirement for live citation.
- 3
Strengthen your brand entity
Build consistent mentions across review sites, directories, Wikipedia/Wikidata where eligible, and press. The model recommends brands it has repeatedly seen described the same way.
- 4
Add Organization & Product structured data
Emit accurate Organization, Product, and FAQ JSON-LD so machines can extract your facts unambiguously. Generate and validate it with the free JSON-LD tools.
- 5
Publish an llms.txt and answer-first content
Add /llms.txt pointing to your most citable pages, and lead pages with direct, factual answers ChatGPT can quote verbatim.
- 6
Track your ChatGPT citations
Baseline how often ChatGPT mentions and cites you, and against competitors, so you can measure what your changes move.
Free tools for ChatGPT optimization
How ChatGPT actually retrieves and cites pages
ChatGPT SEO starts with a distinction most guides blur: the model that answers and the crawler that fetches are different systems on different clocks. OpenAI runs three named agents, and only one decides whether you show up in a live answer this week.
| Agent | What it does | Affects live citations? |
|---|---|---|
| OAI-SearchBot | Fetches and indexes pages so ChatGPT search can surface and link them — the retrieval crawler behind the source chips in an answer. | Yes. Block it and you cannot be cited from the live web. |
| ChatGPT-User | Fetches a specific page when a user asks ChatGPT to open a link, or an action or custom GPT calls the web. Per OpenAI, not used for automatic crawling or training. | Yes, for user-initiated fetches of your URL. |
| GPTBot | Crawls content that may be used to train future models. The agent most sites think about, and the one with the least effect on this quarter's answers. | No. It shapes what the model knows from memory. |
When a prompt arrives, ChatGPT first decides whether it needs the web at all. Stable facts are answered from memory with no fetch and no citation. Time-sensitive, product-specific or comparison questions tend to trigger search, and a user can force it. The query is rewritten into searches, results come back from OpenAI's providers (historically leaning on Bing), a handful of pages are fetched and read at answer time, and the answer carries source chips pointing at the pages that supplied each claim.
So there are two ways to appear. A mention from memory reflects how consistently the model saw your brand described before its training cutoff. A citation from retrieval reflects whether OAI-SearchBot can reach you, whether you are in the index, and whether the page states the claim in a form the model can lift. The second moves in weeks; the first moves when the model is refreshed.
What ChatGPT favours when it picks a source
Once your page can be retrieved, it competes with every other fetched page for the sentence that goes in the answer. Four properties decide that contest.
✓Structure the model can parse in one pass
A question-shaped heading, a direct answer in the first sentence under it, detail after, and lists or tables for anything with more than two items. The page that states the fact first wins.
✓Freshness on time-sensitive claims
ChatGPT search prefers recently-updated pages for pricing, versions and anything with a year in it. A visible last-updated date and an accurate dateModified tell the retriever the page is current.
✓Corroboration across independent sources
A claim that appears only on your site is a claim about yourself. The same claim on a review platform, a partner page and a community thread is a fact the model repeats with confidence.
✓Entity clarity
One name, one description, one category, used the same way on your homepage, in your Organization schema and in directory listings. Ambiguity about what you are is the commonest reason a brand is skipped for a competitor the model can describe cleanly.
Buried answer
<h2>Our approach to returns</h2> <p>At Acme we believe shopping should be stress-free, which is why we designed a returns experience around you...</p> <p>...</p> <p>Returns are accepted within 30 days.</p>
Answer-first
<h2>What is Acme's return policy?</h2> <p>Acme accepts returns within 30 days of delivery for a full refund; items must be unworn with tags attached. Refunds reach the original payment method within 5 business days.</p>
The second version can be quoted verbatim as the answer to its heading. The first makes the model read to the end and infer, and under a retrieval time budget it will usually cite a competitor that did not make it work that hard.
How to rank in ChatGPT
There is no ranking in ChatGPT in the ten-blue-links sense; you are one of the handful of sources an answer cites, or you are absent. This order finds problems cheapest first.
- 1
Prove OAI-SearchBot can fetch your key pages
Check robots.txt, then check what your CDN returns to a request with the OAI-SearchBot user agent — a 403 from a managed rule looks identical to a Disallow. Fetch the page with no JavaScript and confirm the body text is in the HTML.
- 2
Get into the index ChatGPT searches
Verify the site in Bing Webmaster Tools, submit the sitemap, and clear indexing errors. A page ChatGPT search cannot find in its providers' results is never fetched, however good it is.
- 3
Write the prompts before the pages
List the questions buyers type: the category question, "best X for Y", the comparison against each competitor, "is X worth it". Map each to the one page that should answer it, and note the ones with no page at all.
- 4
Make each target page answer-first
Put the question in the heading and a self-contained answer in the first sentence below it. One claim per passage, true without surrounding context, because that is how it will be quoted. Brand story goes below the answer.
- 5
Ship structured data and an llms.txt
Organization on the homepage, Article with a real dateModified, FAQPage where you answer questions, Product where you sell. Validate it. Publish /llms.txt pointing at the pages worth quoting.
- 6
Earn the same description off-site
ChatGPT names brands it has seen described consistently by others. Get your homepage's one-line description repeated on review platforms, directories, partner pages and the community threads where your category is discussed. This is the slow lever that moves memory-based mentions.
- 7
Re-run the prompt set and compare
Run the same prompts on a fixed cadence and record, per prompt, whether you were mentioned, cited, and who appeared instead. Retrieval citations should move within weeks of a fix; memory mentions wait for a model refresh. Knowing which is which stops the team calling a change a failure too early.
Accidental block
User-agent: * Disallow: / User-agent: Googlebot Allow: /
Retrieval allowed
User-agent: OAI-SearchBot Allow: / User-agent: ChatGPT-User Allow: / # training crawler — your call User-agent: GPTBot Disallow: /
The first file was written to keep a staging site private and shipped to production. Googlebot still works, so nobody noticed, but every OpenAI agent falls through to the wildcard and is blocked. The second makes a separate, deliberate decision for retrieval and for training.
ChatGPT SEO checklist
The short version, for the page you are working on right now.
✓OAI-SearchBot and ChatGPT-User allowed
In robots.txt and at the CDN, with a real fetch to prove it.
✓Indexed in Bing
Site verified, sitemap submitted, no coverage errors on the target URLs.
✓Body text present without JavaScript
The claim you want quoted is in the server-rendered HTML.
✓Question heading, answer first
The first sentence under each H2 answers the H2 on its own.
✓Schema validated, dates accurate
Organization, Article with dateModified, FAQPage or Product where they apply.
✓Same one-line description off-site
Reviews, directories and community threads describe you the way your homepage does.
Measure it: are you actually being cited?
ChatGPT does not send a referral header that separates a citation click from a typed URL, and the answer itself is invisible to analytics. The only reliable measurement is to ask the engine the questions your buyers ask and record what comes back.
AI search monitoring runs a fixed prompt set against ChatGPT on a schedule and records, per prompt, whether you are mentioned, whether one of your pages is cited, the sentiment, and which competitors and third-party domains appear instead. Because the prompt set is fixed, a rewrite or a robots.txt fix shows up as a change in citation share rather than an anecdote.
Read retrieval and memory separately. A citation is retrieval and responds to on-page and access fixes within weeks; a mention with no citation is memory and responds to off-site consistency over months. One combined number hides which lever is working.
Frequently asked questions
How do I get my brand mentioned in ChatGPT?+
Make sure OpenAI's crawlers can reach you, get indexed in Bing, build consistent brand mentions across authoritative sources, and add structured data plus answer-first content. ChatGPT recommends brands it repeatedly encounters described the same way.
Does ChatGPT use Google or Bing?+
ChatGPT search has historically relied on the Bing index for live web results, not Google. That makes Bing indexing a practical prerequisite for being cited in real time.
What is the difference between GPTBot and OAI-SearchBot?+
GPTBot crawls content that informs the trained model; OAI-SearchBot retrieves pages for ChatGPT search results. Allow both in robots.txt to be eligible for training and live citation.
Is ChatGPT optimization the same as SEO?+
They overlap but differ. Classic SEO targets ranking in search result pages; ChatGPT optimization targets being named or cited inside an AI answer, which weighs brand-entity strength, crawler access, and answer-shaped content more heavily.
How do I measure ChatGPT visibility?+
Use an AI visibility tracker to monitor how often ChatGPT mentions your brand for your key prompts, which sources it cites, and how you compare to competitors over time.
What is ChatGPT SEO?+
ChatGPT SEO is the practice of making your pages retrievable, quotable and corroborated so that ChatGPT cites or names your brand in its answers. It covers OAI-SearchBot access, Bing indexing, answer-first page structure, validated structured data and consistent off-site descriptions. There is no ranked list to climb; you are one of the few cited sources or absent.
How long does it take to rank in ChatGPT?+
Citations from live retrieval can change within days to weeks of fixing crawler access or restructuring a page, because ChatGPT search fetches at answer time. Mentions from the model's memory only change when the model is refreshed, so off-site brand consistency is a months-long investment. Track the two separately.
Should I block GPTBot but allow OAI-SearchBot?+
You can, and many publishers do. GPTBot only affects whether your content may be used to train future models; OAI-SearchBot is what fetches and links your pages in ChatGPT search. Blocking GPTBot has no effect on live citations. Make the two decisions separately in robots.txt and confirm your CDN is not overriding either.
Does ChatGPT-User respect robots.txt?+
ChatGPT-User is the agent used when a user asks ChatGPT to open a specific page, or an action or custom GPT fetches the web on their behalf. OpenAI states it is not used for automatic crawling or training. Treat it as a user-initiated fetch: if you want links to your site readable inside ChatGPT, do not block it.
See how ChatGPT cites your brand today
Run a free scan to baseline your visibility, then fix the gaps.
Run a free scan →