Why isn't ChatGPT citing your website?

You ask ChatGPT about your own topic, watch it answer with confidence, and notice it cites three competitors and never you. You know your page is better. It is still invisible. That gap has a small set of causes, and every one of them is something you can check this afternoon.

Why this happens

ChatGPT pulls names two different ways, and you need both working.

The first is live web search. When ChatGPT browses, it sends a crawler called OAI-SearchBot to fetch current pages, then writes an answer grounded in what it just read. This is where most citations are won, because it reflects the live web rather than a frozen snapshot.

The second is training data, gathered by GPTBot. That is the corpus baked into the model at its cutoff date. If your brand was repeated across enough sites before that cutoff, the model "knows" you even with browsing off.

So when ChatGPT won't name you, it is usually one of four things:

1. You block the wrong bot. Your robots.txt disallows OAI-SearchBot or GPTBot, often by accident, often left over from a "block all AI" rule someone added in a panic. If the crawler can't fetch the page, it can't cite it.
2. You're not in the retrieval set. ChatGPT's search step runs your question against an index and pulls back a handful of candidates. If you don't rank or aren't indexed for that query, you never enter the pool it reads from. No retrieval, no citation.
3. The content isn't extractable. The bot fetched the page but couldn't lift a clean fact from it. The answer is buried, the text is actually an image, or the whole body only appears after JavaScript runs and the crawler saw an empty shell.
4. Nothing else on the web backs you up. Models lean toward facts repeated across several independent sources. This is how retrieval-augmented generation hedges its bets: a claim only your site makes is riskier to attribute than one three trusted sites agree on. If nobody else mentions you, you look uncorroborated.

Most "ChatGPT ignores me" cases are one or two of these, not all four. The point of the checks below is to find out which.

Check this first

Do these in order. They take about ten minutes and they tell you where the problem actually lives.

Read your own robots.txt. Go to yoursite.com/robots.txt and search the text for GPTBot and OAI-SearchBot. A line like User-agent: GPTBot followed by Disallow: / means you have shut OpenAI out. These are two separate decisions: you can block GPTBot to keep your content out of training while still letting OAI-SearchBot retrieve and cite you. If you want to be cited, that second bot must be allowed.

Ask ChatGPT directly, with web search on. Turn on browsing and ask it the question your customers would ask. "What's the best tool for X?" or "How do I do Y?" Then read who it cites. If it names competitors and not you, you now know the answer exists in its retrieval set and you are simply not in it. That is a ranking and corroboration problem, not a robots one. The same dynamic plays out on other engines, and the breakdown in why Perplexity cites your competitors but not you maps cleanly onto ChatGPT.

Check whether your page is server-rendered. Open the page, right-click, and choose "View Page Source" (not "Inspect", which shows the rendered DOM). Search that raw HTML for a sentence from your main content. If your text is there, the crawler can read it. If the source is mostly empty <div> tags and your content is nowhere to be found, it only appears after JavaScript runs, and OAI-SearchBot likely sees nothing.

Search the wider web for your brand. Run a plain Google search for your company plus your main claim. If the only result is your own site, there is nothing for a model to corroborate against. That is the hardest gap to close, and the most telling.

The fixes

Work top to bottom. The early ones are quick and unblock everything after them.

1. Allow the right bots. If your robots.txt blocks OAI-SearchBot, remove that disallow. Keep blocking GPTBot if you genuinely object to training, but understand that is a separate choice from being cited. The retrieval bot is the one that puts you in answers. While you're in there, make sure you aren't accidentally blocking Googlebot or Bingbot either, because ChatGPT's retrieval leans on conventional search indexes underneath.

2. Get indexed and ranked for the query. Retrieval starts from a search step, so the old fundamentals still decide whether you make the shortlist. Be in the index, have a page genuinely about the question, and earn enough relevance to surface in the top results. There is no AI-specific shortcut around this part. If you don't rank, you don't get retrieved.

3. Write extractable answers. Give the model a clean sentence it can quote and stand behind. Three habits do most of the work:

• Lead with the answer. State the fact in the first sentence or two of the relevant section, complete and standalone. "Refunds are available within 30 days of purchase" can be lifted as-is. A paragraph that circles the point for four sentences cannot.
• Use question-shaped headings. A heading like ## How long do refunds take? mirrors how people ask and how the retrieval step matches. The answer sits right underneath.
• Keep the real content in the HTML. Server-render anything that matters. Don't lock prices, specs or policies inside images or PDFs. If you can't select the text with your cursor, assume the crawler can't read it.

4. Earn third-party mentions. This is the slow one and it matters most for corroboration. When other sites that the model already trusts mention you, RAG has independent sources to cross-check, and citing you stops being a risk. Guest pieces, being quoted in roundups, original data others reference, a listing in the directories your field actually reads. You can't fake this, and it's the difference between a model knowing you exist and trusting you enough to put your name in print. The same corroboration machinery is also why models repeat outdated facts about you; if ChatGPT is naming you but getting the details wrong, correcting what ChatGPT says about your business is the companion problem.

The robots fix can land today. Extractable answers take an afternoon per page. Corroboration is a quarter of patient work, but it's the moat. For the full playbook across every engine, the companion piece on getting cited by ChatGPT and Perplexity and the guide to getting found by AI search go deeper than this one can.

A faster way to find the gap

Running these checks by hand, for one page, against one engine, is doable. Doing it across ChatGPT, Perplexity, Claude and Google AI Overviews, and re-checking after every site change, is the part nobody keeps up with. Rankport's AI visibility checker runs exactly the diagnostics above automatically: it reads your robots.txt for each bot, confirms your pages are server-rendered and extractable, and shows which engines are citing you and which aren't, in plain English. If ChatGPT is skipping you, it tells you which of the four reasons is to blame.