Does llms.txt Actually Work?
The honest answer
Mostly no — and partly yes, for a narrower job than the hype claims. llms.txt does not influence search rankings, does not make AI assistants cite you, and is barely fetched at all today. What it does is cheap, useful packaging for the agents that do read it — and for that job it still earns a place in your root directory.
A scanner that checks llms.txt owes you the unvarnished version, so here it is, sourced.
What the data says
Ahrefs measured 137,210 domains with live traffic in May 2026 and classified every request to their /llms.txt files. 97% of the files received zero requests — nothing fetched them at all. Among the ~3% that got any traffic, 96% of requests came from bots of all kinds, and all AI-bot categories combined accounted for about a fifth of that already-tiny volume.
That matches what the vendors say — or rather, don't: no AI vendor has published a statement that its crawlers consume third-party llms.txt. The companies most associated with the file — Mintlify, Yoast, Rank Math, Shopify — are all generators of it, not consumers.
Google's position
Google has been unusually direct. In April 2025, John Mueller compared llms.txt to the keywords meta tag: "none of the AI services have said they're using llms.txt — and you can tell when you look at your server logs that they don't even check for it." In June 2026 he called its value "purely speculative for now." That month, Google's Search Central changelog added formal guidance: Google Search does not use llms.txt — it neither helps nor harms rankings.
And yet, the same company's Lighthouse now ships an Agentic Browsing audit category that checks your llms.txt. The two positions aren't contradictory — they draw the same line we do: llms.txt is not a search signal; it is agent infrastructure.
What llms.txt is still good for
- The agents that fetch on demand. A user-directed agent landing on your site can grab one file and learn what you offer, which endpoints exist, and where the machine-readable surfaces are — instead of crawling and guessing. That works today; it just isn't proactive consumption.
- A cross-reference hub. The highest-value lines in an llms.txt point at your MCP endpoint, OpenAPI spec, and payment discovery files — things an agent cannot reliably find otherwise. This is what AgentGrade's llms.txt checks actually score.
- It costs ten minutes. The asymmetry matters: near-zero cost, real (if narrow) benefit, and free insurance against the ecosystem maturing around it.
What we score — and don't
AgentGrade scores llms.txt as an agent-usability signal: does it exist, does it parse, does it give an agent navigable structure and cross-references. We have never scored it as an SEO or citation lever, because there is no evidence it is one. If a vendor sells you llms.txt as a rankings play, the numbers above are your counterargument.
FAQ
Does llms.txt help SEO?
No. Google's documented position is that Search does not use it — it neither helps nor harms. Its purpose is agent usability, not rankings.
Do ChatGPT or Claude read llms.txt?
Neither OpenAI nor Anthropic has published a statement that their systems fetch third-party llms.txt. Both companies publish llms.txt files for their own documentation sites — which tells you they consider the format useful packaging, not that their crawlers consume yours.
Should I still add llms.txt?
Yes — because it takes minutes, helps the on-demand agents that do read it, and is the natural home for cross-references to your MCP endpoint, OpenAPI spec, and payment files. Validate it with the llms.txt validator.
Why does AgentGrade check it if nobody fetches it?
Because our score measures whether an agent that arrives can use your site, not how many arrive. A parseable llms.txt with real cross-references measurably shortens an agent's path to your capabilities — that's the claim we score, and the only one.