Does llms.txt actually affect AI citations?

llms.txt has become one of the more talked-about ideas in AI SEO, so it’s worth asking whether it actually does anything for your visibility in AI answers. The proposal, introduced in September 2024 by Jeremy Howard of Answer.AI, is a markdown file placed at yoursite.com/llms.txt that curates a site’s most important content for large language models — the idea being to hand an LLM a clean map of your site rather than making it wade through cluttered HTML.

The problem is that the engines most people care about don’t appear to use it. None of the major providers — OpenAI, Anthropic, Google, Meta, or Mistral — has publicly committed to using llms.txt as a signal in their production search or answer surfaces, and Google has actually said no on the record: in July 2025 Gary Illyes confirmed that Google doesn’t support the file and isn’t planning to, while John Mueller compared it to the long-discredited keywords meta tag. That skepticism is borne out by crawler behaviour, too — one analysis of more than 500 million AI-bot visits over a 90-day window found that only 408 of them requested llms.txt directly, and overall adoption sits at roughly 10% of sites according to an SE Ranking study of 300,000 domains.

None of this means the file is useless, but its real value lies somewhere other than citations. The strongest use case today is with AI coding assistants rather than general answer engines: tools like Cursor, Claude Code, GitHub Copilot, and Windsurf look for /llms.txt when they are pointed at a documentation site, so a business publishing developer docs may find it genuinely helpful for those tools. That is a different job from earning a citation in ChatGPT or Perplexity.

So if you are deciding whether to bother, the honest answer is that llms.txt is a low-cost, low-risk addition rather than a priority. It will not hurt anything, and it can help coding assistants read your documentation, but if your actual goal is to be cited in AI answers, your effort is far better spent on the things engines demonstrably rely on — allowing AI crawlers through in robots.txt, keeping your page structure clean, and adding structured data.