The truth about llms.txt — what the data actually shows

What it is

llms.txt is a community-proposed plain-text file placed at the root of a website that offers AI systems a structured summary of the site. The idea sounds sensible — a business card written for machines.

But an idea sounding sensible is not the same as machines actually reading it. So instead of repeating the pitch, here is what the measurement shows.

What the data shows

• 97% of published llms.txt files are never fetched. A June 2026 study analyzed 137,210 domains: of the sites that published the file, 97% saw zero requests to it.
• Google says Search ignores it. Google's own developer documentation states you don't need to create machine-readable AI files to appear in AI Overviews or AI Mode — “Google Search ignores them.”
• Neither OpenAI nor Anthropic documents reading it. Their published crawler documentation names robots.txt, feeds, and ordinary crawlable pages — never llms.txt.
• The few real fetches come from coding assistants reading developer documentation — not from answer engines deciding which plumber to recommend.

Any vendor selling “llms.txt deployment” as an AI-visibility fix is selling you a file that, on the best available evidence, nothing you care about will ever read.

Why we publish one anyway

You'll find ours at signalflair.ai/llms.txt. It costs nothing to serve, it's honest, and if the standard ever earns real adoption, we're ready. That's the right size for this file: a free courtesy — not a line item on your invoice.

If we sold it to you as a visibility fix, we'd be selling the myth. Our whole product is that we don't do that — every finding we hand you is backed by evidence you can check.

What actually moves AI visibility

• Answer-crawler access. The one fix with peer-reviewed causal support: sites blocking the crawlers that feed AI answers measurably lose AI visibility (SIGIR 2026, 11,500 real user queries). Blocks are usually a security plugin's default — nobody decided them.
• Ordinary crawlable pages. Google's stated requirement for AI features is simply being indexed and snippet-eligible. The engines read your actual site.
• A consistent Google Business Profile. Google documents that it can update your profile from what the rest of the web reports — and that you can't manage all Google updates. Inconsistent facts don't just confuse AI; they invite Google to rewrite you.
• Consistent facts everywhere AI cross-references — your site, directories, and reviews telling one story, machine-readably.

That's the work Signal Flair does — and every finding carries the fingerprint of the file it came from, so you can verify us the way AI verifies you.

How it compares to robots.txt & schema

• robots.txt — controls which crawlers are allowed in. This one is load-bearing: it decides whether answer engines can read you at all.
• Schema markup — machine-readability hygiene that makes your facts legible and earns rich results. Useful; not a citation vending machine.
• llms.txt — a proposal almost nothing reads today. Publish one if you like; pay for one never.