Does llms.txt Help With ChatGPT Citations? An Honest Answer

No. As of mid-2026, adding an llms.txt file to your site does not get you cited in ChatGPT, and there is no public evidence it helps with ChatGPT citations. ChatGPT's web crawlers do not request the file when they browse, and Google has said plainly it ignores it. The file's real, narrow value is in agentic and IDE tools, plus Perplexity-style assistants that fetch a page directly.

What is llms.txt and what was it supposed to do?

llms.txt is a proposed Markdown file you place at the root of your domain (like yoursite.com/llms.txt). The idea, proposed in 2024, is to give large language models a clean, curated map of your most important pages — titles, short descriptions, and links — so an AI reading your site does not have to wade through navigation, ads, and cookie banners.

It is a sensible idea on paper. A clean index of "here is what matters on this site, in plain Markdown" is exactly the kind of thing a model would love to read. The hope many people attached to it was bigger: write this file, and ChatGPT, Google's AI Overviews, and Perplexity will start citing you more.

That second part is where the myth lives. There is a difference between "a model could read this if pointed at it" and "search-time AI crawlers go fetch this file and use it to decide who to cite." Those are not the same thing, and the evidence on the second one is not good. The honest framing matters here, because chasing llms.txt for ChatGPT citations means time not spent on things that actually move the needle.

Does ChatGPT actually read llms.txt for citations?

There is no public evidence that it does. When ChatGPT search browses the live web to answer a question, its AI crawler fetches the actual page that ranks for the query — the real HTML — not a special llms.txt index. Server logs shared across the SEO community consistently show AI crawlers requesting normal pages, sitemaps, and robots.txt, while llms.txt requests are rare to nonexistent.

The mechanism explains why. ChatGPT, like other answer engines, mostly relies on a search layer (its own index plus partners) to find candidate pages, then reads those pages. A curated file at /llms.txt sits outside that retrieval path. Nothing in the pipeline is wired to go look for it, so writing one does not insert you into the set of pages the model considers.

This does not mean llms.txt is fraudulent or useless — it means it is not a ChatGPT citation lever. If your goal is being quoted in ChatGPT answers, the file is not the tool. The tools are well-structured pages, clear answer-first content, and being genuinely citable, which we cover in the GEO guide.

Does Google support llms.txt?

No. Google has been unusually direct about this. Google's John Mueller has publicly compared llms.txt to the old keywords meta tag — something sites fill in that the search engine simply does not use — and Google has said it does not use the file. Googlebot already renders and understands normal HTML pages, so there is nothing for llms.txt to add on Google's side.

So for Google Search and AI Overviews, you can treat llms.txt as having no demonstrated ranking or citation value today. That is not a prediction about the future — standards can get adopted — it is the state of things now, based on public statements from Google's own people. Building a content strategy around a signal the dominant search engine has openly disavowed is a bad bet.

So where does llms.txt actually help?

It helps where a tool fetches your file on purpose, by name. That is a real and growing layer, just not the search-citation layer most people are aiming at.

The strongest real use is documentation and developer tools. When a developer is using a coding agent and points it at your docs/llms.txt, the file is genuinely useful — it is a clean, token-efficient map the agent reads directly. Libraries and SaaS products with API docs get real value here. That is a different job than "rank me in ChatGPT," and it is the job llms.txt is actually good at.

What should you do instead to get cited in AI search?

Optimize the page itself, because the page is what the AI reads. Answer Engine Optimization and Generative Engine Optimization come down to making real HTML pages that are easy to quote and easy to trust.

Concretely, the moves that have evidence behind them:

• Answer-first writing. Put a complete, self-contained answer in the first 40–80 words of a page, so a model can lift it cleanly. Princeton's 2024 GEO study ("GEO: Generative Engine Optimization") found that adding citations, quotations, and statistics was associated with a higher likelihood of being included in AI answers — structure and credibility win, not a sidecar file.
• Clean, crawlable HTML. Make sure your real pages render without heavy client-side gymnastics, and that your robots.txt allows the AI crawlers you want.
• Schema and FAQ markup. Structured data helps machines parse what your page asserts.
• Be genuinely referenced. Mentions and links from places AI already trusts feed the index that ChatGPT and Perplexity draw from.

If you are weighing the wider picture of whether AI search rewrites the rules, the AEO playbook for B2B websites and the GEO guide go deeper. In our experience, a single well-structured, answer-first page does more for AI citations than any number of llms.txt files.

Should you add llms.txt anyway?

You can, and it is low-risk — but be honest with yourself about why. If you ship developer docs and want agents to read them well, yes, add it; it is cheap and genuinely helpful there. If you are adding it purely hoping for ChatGPT or Google citations, you are spending effort on a signal those engines do not currently use.

The pragmatic rule: maintain llms.txt as a courtesy and a docs aid, keep it small and accurate, and never let it sit at the top of your AI-search to-do list. The page is the product. The file is a bonus.

FAQ

Does llms.txt help with ChatGPT citations?

There is no public evidence that it does. ChatGPT's web search reads the actual pages that rank for a query, not a special llms.txt index, so the file sits outside the path it uses to pick citations. To get cited, optimize the page itself.

Does Google use llms.txt?

No. Google (via Search Advocate John Mueller) has publicly said Google does not use llms.txt, comparing it to the unused keywords meta tag. It has no demonstrated ranking or AI Overviews value in Google today.

Which AI tools actually read llms.txt?

Mostly agentic and IDE tools — developers point coding agents like Claude Code and Cursor at a project's llms.txt for clean documentation. Perplexity can also use it as context when fetching a specific URL. Search-time crawlers for ChatGPT and Google generally do not.

Is it worth adding llms.txt to my site?

It is worth it if you publish developer docs and want AI agents to read them cleanly, since it is cheap and low-risk. It is not worth prioritizing if your only goal is ChatGPT or Google citations, because those engines do not currently use it.

What works better than llms.txt for AI search visibility?

Answer-first content, clean crawlable HTML, schema markup, and being genuinely referenced by trusted sources. Princeton's GEO study found that adding citations, quotations, and statistics to a page was associated with a higher likelihood of appearing in AI answers — page quality beats a sidecar file.