Myths & debunks
Does llms.txt Actually Do Anything for AI Citations?
In 2024, a simple proposal started circulating: add a file called llms.txt to your website. The idea was that AI language models would read it, understand your business, and represent you more accurately in their answers.
It spread quickly. By 2025, it had become a routine checklist item — something to add before anyone asked why you hadn’t. A real experiment ran the numbers.
Of 62,100 AI bot visits tracked over 90 days, 84 of them hit the llms.txt file. That’s 0.1%.
Does llms.txt change AI citation rates?
No. SE Ranking analyzed 300,000 domains and found a 10.13% adoption rate — and no detectable difference in citation rates between domains with and without the file.
The number makes sense when you understand what AI systems are actually doing. When ChatGPT, Claude, or Gemini generate an answer that mentions a business, they’re drawing on what their training corpus — or their live search crawl — has already established. They aren’t querying a file at the moment of response. llms.txt describes a business to a system that has usually made up its mind before visiting your site.
The spec’s creator, Jeremy Howard, designed llms.txt for a specific use case: helping AI coding assistants and documentation tools navigate technical sites more efficiently. That’s a real problem worth solving. Whether your AI coding assistant reads your llms.txt when pulling in a library reference is different from whether ChatGPT names your business when someone asks for a service recommendation.
Will search engines ever support it?
Not soon, according to Google. Gary Illyes of Google Search Central has stated publicly that Google has no support for llms.txt and no current plans to add it.
This matters because Google is one of the primary data sources large AI systems draw from — via Bing, via the open web, via the training data that includes years of Google-indexed content. If the search engine that feeds the largest AI answer systems doesn’t read the file, the file doesn’t reach the audience you’re trying to reach through that channel.
What actually determines whether AI recommends your business
The answer is less satisfying but more actionable than adding a file.
Ahrefs studied 75,000 brands and found that brand mentions in trusted web sources — directories, review platforms, local news, industry sites — correlate 0.656 to 0.709 with AI citation share. YouTube brand mentions correlate 0.737. Traditional backlinks, the metric most SEO tools optimize for, correlate weakly.
The mechanism: AI engines recite what their training corpus established as fact, weighted by how often they’ve seen it from trusted sources. A business with a Yelp listing, a Nextdoor mention, 40 Google reviews, and a clean directory entry is more citable than one with a well-crafted llms.txt and nothing else.
llms.txt is not harmful. If your site’s technical team can add it in 20 minutes, there’s no reason not to. But if you’re trying to change whether AI engines name your business, the files that need attention are your directory listings — not the one in your website root. The AI Visibility Myths cluster covers the other common misconceptions that follow from this same pattern.
When we ran an AI visibility audit on our own business, we were cited in 0 of 100 engine responses for non-branded queries. Our llms.txt wasn’t the problem. Here’s what actually was.
→ We scored ourselves 0/100 on citations. Read the full audit.
→ Get a free AI Visibility Snapshot — we’ll map the actual gaps in your citation footprint.
Data from Otterly.ai’s llms.txt experiment and SE Ranking’s 300,000-domain analysis, 2024–2025. Gary Illyes statement via Google Search Central. AI crawler behavior and platform support for new standards change frequently — if you’re reading this more than six months after the publication date above, check for updates.