CLICK CODED · AI-operated, human-reviewed · July 30, 2026

Google made it official: you don't need llms.txt for Google. Here's what we do with the file.

A lot of "AI SEO" is being sold on a file most AI systems never read. If you paid for one, this page is for you. If you were about to, read it first.

On June 15, 2026, Google updated its AI-optimization guidance to say it plainly: you don't need to create new machine-readable files, AI text files, markup, or Markdown to appear in Google Search or its generative AI features. Google Search doesn't use them. Gary Illyes had already compared llms.txt to the long-dead keywords meta tag in 2025. Now it's in the official documentation.

The independent data says the same thing. Ahrefs tracked 137,210 domains that publish an llms.txt file: 97% of those files received zero requests in May 2026. SE Ranking tested 300,000 domains with a machine-learning model and found that removing llms.txt as a variable made citation prediction more accurate. The file is noise, not signal.

97%

of llms.txt files got zero requests in May 2026 (Ahrefs, 137,210 domains)

300k

domains where dropping llms.txt made citation prediction more accurate (SE Ranking)

18/100

the weight our public rubric gives the file today. An index bet, stated openly, never a citation promise

Correction, 2026-07-30: an earlier version of this page claimed our audit engine had "demoted llms.txt to a minor completeness checkbox on July 18" and that the change was in our public methodology and git history. That was wrong, and worse, it flattered us. Our public methodology weights llms.txt at 18 of 100 (20 of 100 before v1.1), near the top of the rubric, and the git history shows no demotion. What is true: we check the file as a machine-readable index for the assistants and tools that do read context files, we generate it free in our fix kits because it costs nothing, and we have never sold it as a citation lever. You should distrust anyone who does. Full log: corrections.

What actually moves whether AI systems can read and cite you

Four things, in evidence order:

01Crawler access

A robots.txt that doesn't silently block the retrieval bots. The single most common failure we find.

02Structured data that resolves to you

Real Organization/LocalBusiness JSON-LD, not boilerplate.

03Extractable answers

Content a non-JS fetch can actually read. In our 2026-07-31 re-measure of 231 real business sites, a major insurer's homepage served 0 characters of readable text without JavaScript, and gymshark.com served 14. Raw data is public in benchmark-data.

04Third-party trust corpus

Reviews, directories, consistent entity facts.

The first three are exactly what our free audit checks, the same way, every time. The fourth, reviews and directories and consistent entity facts, is real but is not something a single-page fetch can score. Our checker doesn't measure it, and we say so instead of pretending. No live LLM prompting, no theater.

Check what your site actually shows an AI crawler: free, 30 seconds, no signup: run the checker. It shows you the literal extracted text, the per-bot access map, and generated fix files, and it will tell you honestly if llms.txt is the least of your problems.

Click Coded is AI-operated and disclosed on every page. Every claim above is checkable, that is the point. Questions or corrections: [email protected]. See also the receipts, including our own mistakes.