# Why ChatGPT doesn’t cite your site: a diagnostic list

Source: https://citetune.com/blog/why-chatgpt-doesnt-cite-your-site

Written by: CiteTune Team · Published: 2026-10-03 · Updated: 2026-10-03 · 8 min read

## Short answer

ChatGPT usually fails to cite a site for a technical reason before a content reason. Work down six layers in order: bot access, indexing, text in the raw HTML, a quotable answer, a clear identity, and an answer that links back. Fix the first failing layer, then re-measure.

## First, confirm you are really not cited

Before diagnosing, make sure the problem is real. An answer engine does not return the same answer to the same question every time, and a single screenshot from one person’s account is not a measurement. We state this on our own [surfaces page](https://citetune.com/surfaces): no guarantee that two runs of the same prompt return the same answer, and every sample should report how many prompts were attempted and how many completed.

Separate two outcomes that look alike. **Retrieval** is the engine reading your page. **Citation** is the engine naming or linking you in the answer. A page can be retrieved and never cited, and a brand can be mentioned without a link. The difference is explained in [retrieval versus citation](https://citetune.com/glossary/retrieval-vs-citation), and our [methodology](https://citetune.com/methodology) page shows how we count each.

## The six layers, in the order to check them

Each layer depends on the one above it. A perfect FAQ block does nothing if a firewall refuses the crawler, so start at the top and stop at the first failure.

| Layer | Question | Fast check | CiteTune rules |
|---|---|---|---|
| 1. Access | Can the bots fetch your pages? | robots.txt plus a live request per bot | robots_blocks_all_agents, robots_blocks_ai_agent, site_blocks_crawlers |
| 2. Discovery | Can engines index and canonicalize the URL? | noindex, canonical, redirects, sitemap | page_noindex, canonical_off_site_or_broken, redirect_chain_or_loop, sitemap_lists_bad_urls, sitemap_missing_high_value_pages |
| 3. Readability | Is the text in the first HTML response? | View source, or fetch without JavaScript | raw_rendered_divergence |
| 4. Answer fit | Does the page state a quotable answer? | Read the first screen as a stranger | thin_evidence_page, faq_coverage_gap, conflicting_title_h1 |
| 5. Identity | Is it clear who you are? | Organization markup, consistent name | missing_or_invalid_organization_jsonld, identity_mismatch |
| 6. Attribution | Does the answer link back? | Compare mentions with cited URLs | mention_not_cited_blocked, mention_not_cited_no_canonical, mention_not_cited_missing_url |

> Not every rule in the table runs in the free scan. The three mention_not_cited rules need stored AI answers and are built but switched off in production. The rule faq_coverage_gap needs topic clusters from tracked prompts, which the free scan does not have. The rule raw_rendered_divergence needs a rendered snapshot, which the free scan does not take. The remaining rules in the table read your site, and the free scan raises them from raw HTML. See [mentioned but not cited](https://citetune.com/blog/mentioned-but-not-cited-ai-answers) for the attribution rules.

## Layer 1: can the bots reach you?

OpenAI documents that sites opted out of OAI-SearchBot are not shown in ChatGPT search answers, apart from possible navigational links; GPTBot is a separate training crawler. Check the relevant robots.txt rule, then verify real crawler access if a test request using a bot user agent returns 403, since IP-based bot verification can differ.

Run the free [AI bot access checker](https://citetune.com/tools/ai-bot-access-checker) for the per-bot result. For the full explanation and the Cloudflare settings that most often cause a silent 403, read [is your site blocked from GPTBot](https://citetune.com/blog/is-your-site-blocked-from-gptbot).

## Layer 2: can engines find and index the URL?

ChatGPT search is not a pure crawl of your site. As of our 30 September 2026 reading, OpenAI’s help center said it also draws on partner content and third-party search providers. We cannot see which provider serves a given answer, so treat your standing in classic indexes as a prerequisite and check Google Search Console and Bing Webmaster Tools both.

- A stray noindex tag or X-Robots-Tag header on a page you want cited.
- A canonical tag that points to a filtered, campaign or off-site URL.
- Redirect chains and loops. Vercel measured that 14.36 percent of GPTBot’s fetches were redirects and 34.82 percent of ChatGPT crawler fetches landed on 404 pages, against 8.22 percent for Googlebot (one month of data, published 17 December 2024).
- A sitemap that lists dead or redirected URLs, or leaves out the pages that answer questions.

Those numbers are older and from one platform’s traffic, so read them as direction, not as a benchmark for your site.

## Layer 3: is the answer in the HTML the bot receives?

Vercel’s 2024 analysis found that the AI crawlers it measured, including OpenAI’s three bots, ClaudeBot and PerplexityBot, did not render JavaScript, while the Google and Apple bots in its test did. If your headline, price or product description appears only after a script runs, those crawlers may miss it. That is a risk for client-rendered React, Vue and Angular sites; see [our guide to testing and fixing content that AI crawlers may miss](https://citetune.com/blog/react-spa-invisible-to-ai-crawlers).

## Layer 4: does the page contain a quotable answer?

Here the bot can read the page and still has nothing to use. Common patterns:

- Prices, shipping times or return terms live in images, tabs loaded by script, or PDFs.
- The answer sits under a long brand introduction instead of at the top.
- The title says one thing and the H1 another, so the topic is ambiguous.
- Structured data disagrees with visible text, for example a different price. Google’s guidance is that structured data must match the visible text.

The remedy is plain editing: put a short, self-contained answer first, write numbers and conditions as text, and align the title and H1. CiteTune flags thin pages, missing FAQ coverage and title and H1 conflicts, and for each it can draft a reviewable change rather than a rewrite you have to trust.

## Layer 5: does the engine know who you are?

If your brand name is a common word or is written three ways across your site, profiles and marketplaces, an engine may not connect the pages to one entity. This layer is a risk area, not a proven cause, and we flag it with lower confidence than a blocked bot. The cheap fixes are consistent naming, a clear About page and an Organization record on the homepage with matching name and URL. The details are in [Organization schema and entity consistency for AI answers](https://citetune.com/blog/organization-schema-for-ai-answers).

## Layer 6: mentioned but not linked

If an answer names your brand without a link, the cause may be on your side or the engine’s: the cited page was blocked, it lacked a canonical URL, or the answer carried no URL for it. CiteTune’s mention rules look at stored answers from tracked prompts, so they apply to monitored projects, not to a one-off free scan. The free scan covers the on-site checks for layers 1, 2, 4 and 5 from raw HTML. It does not render pages, so test layer 3 with the curl check in our JavaScript guide.

## What changes for Google AI Overviews and AI Mode in the US

Google began rolling out AI Mode to everyone in the United States on 20 May 2025, with no Labs sign-up. Google’s documentation says a page must be indexed and eligible to show a snippet to appear as a supporting link in AI Overviews or AI Mode, with no additional technical requirements. For Google, layers 1 to 4 check technical eligibility; they do not guarantee that Google will choose the page as a link.

To see whether you appear, Search Console’s Generative AI performance report counts impressions in AI Overviews and AI Mode by page, country, device and date. Google says it rolled out to all websites worldwide as of 31 August 2026. It shows impressions only, with no clicks or queries, so it tells you whether you are visible and not what the answer said about you.

## What not to spend time on first

Google states that you do not need new machine-readable files or special markup to appear in its generative AI features. Large studies found no link between llms.txt and AI citations (details in [does llms.txt work](https://citetune.com/blog/does-llms-txt-work)). Fix access and readability before either file or markup, because neither repairs a blocked or empty page.

Start with the [free scan](https://citetune.com/scan): up to 30 public pages, no account, and one complete finding with the observed facts quoted in full. Allowing the bots only makes your pages available to them. No tool, including ours, can promise that an engine will cite you.

## Frequently asked questions

### Why does ChatGPT cite my competitor and not me?

Work the layers in order for your own pages first: bot access, indexing, text in raw HTML, a quotable answer near the top. A competitor page that passes all four has an advantage that content volume alone will not fix. Compare a competitor page and yours with the same checks.

### Do I need to apply to be included in ChatGPT search?

OpenAI’s crawler documentation describes no application. It says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, so allowing that bot in robots.txt and at your firewall is the documented step.

### How often should I re-test after making a fix?

Re-run the access check after about a day, since OpenAI says robots.txt changes can take roughly 24 hours to apply. For citations, repeat the same prompt set on a schedule and compare completion rates, because single runs vary.

### Is being mentioned without a link a technical problem?

Sometimes. A blocked page, a missing canonical or an answer that carries no URL can each produce a mention with no link. Other times the engine simply chose not to link. Check the cited URLs and your page state before assuming either.

## Sources

- [OpenAI: Overview of OpenAI crawlers](https://developers.openai.com/api/docs/bots) (accessed: 2026-10-03)
- [OpenAI Help Center: Searching the web with ChatGPT](https://help.openai.com/en/articles/9237897-chatgpt-search) (accessed: 2026-09-30)
- [Vercel: The rise of the AI crawler (17 December 2024)](https://vercel.com/blog/the-rise-of-the-ai-crawler) (accessed: 2026-10-03)
- [Google Search Central: AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) (accessed: 2026-10-03)
- [Google Search Central: Optimizing your website for generative AI features on Google Search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide) (accessed: 2026-10-03)
- [Google: AI Mode rolling out in the U.S. (20 May 2025)](https://blog.google/products/search/google-search-ai-mode-update/) (accessed: 2026-10-03)
- [Search Console Help: Generative AI performance report](https://support.google.com/webmasters/answer/16984139) (accessed: 2026-10-03)
- [SE Ranking: LLMs.txt study (7 November 2025)](https://seranking.com/blog/llms-txt/) (accessed: 2026-10-03)

## Related pages

- [Retrieval vs citation](https://citetune.com/glossary/retrieval-vs-citation)
- [AI crawlers](https://citetune.com/glossary/ai-crawlers)
- [What is AI Mode?](https://citetune.com/glossary/ai-mode)
- [What is an answer engine?](https://citetune.com/glossary/answer-engine)
- [Is your site blocked from GPTBot? robots.txt and 403s](https://citetune.com/blog/is-your-site-blocked-from-gptbot)
- [When React and SPA content may be missed by AI crawlers — and how to fix it](https://citetune.com/blog/react-spa-invisible-to-ai-crawlers)
- [Does llms.txt work? What the evidence says](https://citetune.com/blog/does-llms-txt-work)
- [Organization schema and entity consistency for AI answers](https://citetune.com/blog/organization-schema-for-ai-answers)
- [Are AI agents blocked?](https://citetune.com/tools/ai-bot-access-checker)
- [Free scan](https://citetune.com/scan)
- [Methodology](https://citetune.com/methodology)
