If your site never appears in ChatGPT's answers, the cause is usually one of a small set of technical and content problems — not bad luck. ChatGPT does not rank websites the way Google does. It answers from a mix of training data and live retrieval, and a site can be excluded at several distinct points in that pipeline.

This guide walks through the seven most common reasons, in the order worth checking them. Start with the technical gates, because they are fast to verify and can make every other improvement pointless until fixed.

1. Your site was not in the training data

ChatGPT's base knowledge comes from a training snapshot with a cutoff date. If your site launched or published its key pages after that cutoff, the model will not "know" it from memory. This is the single most common reason for new sites.

The practical fix is not to try to inject yourself into training. It is to make sure you are reachable through ChatGPT's live retrieval paths, which is what the remaining reasons address.

2. AI crawlers are blocked in robots.txt

ChatGPT's operators run several distinct crawlers. GPTBot collects data for model training, while OAI-SearchBot crawls for search-style indexing that can surface in answers, and ChatGPT-User acts when a user explicitly asks the assistant to visit a page. If your robots.txt disallows these user agents, the corresponding path is closed.

A line like this silently removes you from consideration:

User-agent: GPTBot
Disallow: /

Check your file for any Disallow: / rule attached to an AI user-agent. Note that a crawler you do not mention is allowed by default — so the absence of a rule is not the problem. Run the free check at https://ai.ynnadmusic.uk to see exactly which of 19 tracked crawlers your site blocks.

3. You are missing from the search index ChatGPT retrieves from

When ChatGPT browses or searches, it pulls from a search index rather than reading the open web at random. If your pages are not indexed there — because of a noindex tag, a blocking robots.txt rule for the general crawler, a broken sitemap, or low authority — they cannot be retrieved.

Check three things:

  • Your homepage and key pages return HTTP 200 and are not behind a login or paywall.
  • You do not have a global <meta name="robots" content="noindex">.
  • Your sitemap.xml exists, lists canonical URLs, and is submitted to the major search consoles.

4. No llms.txt or other machine-readable signpost

An llms.txt file at your site root is an emerging convention that gives assistants a clean summary of your most important pages. It is not a ranking factor, and there is no public evidence that any major assistant requires it. Its value is convenience: it reduces the work a model must do to find your best pages.

Adding one is free and takes ten minutes. See the llms.txt guide for a copy-paste example. Treat it as a small, low-confidence improvement — not a fix for a blocked crawler.

5. No structured data

Structured data (JSON-LD using schema.org vocabulary) gives machines explicit facts about your pages: what kind of page this is, who wrote it, what questions it answers, and how entities relate. Assistants that parse page content can use these facts to extract clean answers.

For most sites, the highest-value markup is Organization, Article, FAQPage, and BreadcrumbList. A page with a clearly marked FAQ is easier to quote than one where the same question and answer are buried in prose. This is a signal, not a guarantee — structured data improves clarity but does not force citation.

6. Thin, unstructured, or JavaScript-only content

Assistants favour pages that state answers plainly. Content that is padded, vague, or hidden behind heavy client-side rendering is harder to extract and less likely to be cited.

Make the answer to the page's main question visible in the first paragraph. Use descriptive headings. Put important text in the HTML itself rather than injecting it solely through JavaScript. A page that opens with a direct, self-contained answer is far more usable to a retrieval system than one that opens with three paragraphs of background.

7. Few third-party mentions

Answer engines lean on consensus. If reputable sites mention your brand, link to you, and describe you consistently, the model has more confidence that you are a relevant, trustworthy source. A site that exists in isolation — no mentions, no links, no consistent entity description — is easy to overlook.

This is the slowest factor to fix and the one most outside your direct control. Earn mentions through genuinely useful content, digital PR, and consistent naming across your profiles. Avoid manufactured links.

A diagnostic flow you can run today

  1. Check crawler access. Run the free tool and look for blocked AI user-agents. Fix any you did not intend to block.
  2. Confirm reachability. Open the page in a private window, check the status code, and remove any accidental noindex.
  3. Add clarity. Ensure a direct answer appears early, headings are descriptive, and JSON-LD is present.
  4. Add a signpost. Publish an llms.txt and submit your sitemap.
  5. Build mentions. Over months, earn third-party references to the pages you want cited.

Work through these in order. Technical blocks first, then content clarity, then authority. For a broader view, see the 12-point AI visibility checklist.

Frequently asked questions

Does adding an llms.txt get me into ChatGPT?

There is no public evidence that it directly affects inclusion. It is a convenience file that may help an assistant find your key pages, but it does not override robots.txt and is not a ranking factor.

How long does it take to show up in ChatGPT answers?

There is no fixed timeline and no guarantee. Changes to crawl access and indexing can be reflected in retrieval-based answers relatively quickly, while changes to a model's training-based knowledge depend on future training cycles.

If I block GPTBot, am I invisible to ChatGPT everywhere?

No. Blocking GPTBot affects training collection specifically. Search-style retrieval uses different crawlers such as OAI-SearchBot, and explicit user visits use ChatGPT-User. Each is controlled separately in robots.txt.

Is being cited in ChatGPT the same as ranking in Google?

No. They are different systems with different retrieval, ranking, and generation steps. A page can rank well in Google and never be cited by an assistant, and vice versa. Optimize for both, but measure them separately.