DesignRush reports that 97% of llms.txt files on the public web are unread by AI crawlers, a finding that puts paid "AI visibility" packages on notice.

The headline figure lands at a moment when SEO vendors are actively marketing llms.txt deployment as a prerequisite for citation inside ChatGPT, Claude, Perplexity, and Google's AI surfaces. The reporting indicates almost none of those crawlers are asking for the file.

The number is the 97%. That share of llms.txt files deployed across measured sites received no fetch requests from the user agents associated with major AI assistants during the period DesignRush tracked.

What is llms.txt?

• A proposed plain-text file, served at the domain root
• Lists site URLs with short descriptions in a structured markdown format
• Modeled after robots.txt and sitemap.xml conventions
• Intended as a shortcut for LLM-friendly content discovery

The format emerged in 2024 as a proposed standard. Vendors quickly began packaging implementation services and "AI readiness" audits around it.

Why does the 97% matter for operators?

If crawlers do not request llms.txt, the file cannot influence retrieval, citation, or ranking inside any AI assistant. Operators paying monthly fees for "llms.txt deployment and monitoring" are paying for a control surface that the platforms themselves do not consume.

The gap between vendor promise and crawler behavior fits the early phase of any new web standard. Structured data via Schema.org followed a similar curve: published widely for years before crawler adoption caught up.

What the data does not yet prove

DesignRush's headline figure does not include the methodology window, the sample size, or the list of user agents tested. Operators who want to verify the number against their own access logs can do so directly.

• Publish a unique canary URL inside your llms.txt
• Track access logs for 30 days
• Count requests from GPTBot, ClaudeBot, PerplexityBot, and Google-Extended

If the count is zero, your deployment matches the 97% headline. If non-zero, your site falls into the 3%.

Practical takeaways

• Treat "AI visibility" subscriptions that pitch llms.txt as the deliverable with skepticism until vendors publish the fetch data behind the claim.
• Keep robots.txt and sitemap.xml maintained. They still receive crawler traffic at near-universal rates.
• Expect llms.txt to function as a reference document for human reviewers, not as a live signal, for at least the next two quarters.

What to watch next

Two milestones will determine whether the 3% grows.

1. A public commitment from a major AI lab to honor llms.txt directives
2. Inclusion of llms.txt requests in crawler telemetry products from Cloudflare, Vercel, or mainstream analytics platforms

Until either ships, the operator read is straightforward: the file exists, vendors sell against it, and the crawlers, by DesignRush's count, almost entirely do not read it.