One in Eleven Big Sites Has an llms.txt File
- The file is a plain-text guide that points AI models to a site's key pages.
- A study found one on 8.7% of the top 1,000 websites.
- Lighthouse checks for the file, while Google Search ignores it.
- Writing one takes an afternoon, but buyer-focused page text matters more for recommendations.
- Measure AI recommendations with fixed questions, official-domain checks and repeat tests.
A study published August 7, 2026 found that 8.7% of the top 1,000 websites had an llms.txt file. That is 87 of 1,000 sites.
I have built for the web since 1994. I now measure whether AI tools recommend a business, rather than relying on technical checks as a stand-in for that outcome.
The takeaway
Treat llms.txt as a low-cost guide to your site, and judge AI visibility by whether models recommend your business in repeat tests.
What does an llms.txt file tell AI models?
An llms.txt file is a proposed standard: a plain-text file at the root of a site that explains what the site is and points to its important pages. The proposal gives site owners a format for describing a site to AI models.
The file can orient a model before it reaches a page. Its value depends on whether the pages it points to are clear enough to answer a buyer's question. A guide that names a relevant page can help identify where to look, but the page itself carries the explanation a buyer needs.
A study published August 7, 2026 found 51.8% adoption in a panel weighted toward developer sites. The study reports adoption across that panel.
What do Google's guidance and Lighthouse check?
Google Search Central says site owners do not need new machine-readable files for Google Search, and that Google Search ignores them. Google's Lighthouse tool, in its Agentic Browsing audit category, checks whether a site has an llms.txt file. Google's guidance and the Lighthouse audit describe those checks.
The practical conclusion is limited: the file is not needed for Google Search, but Lighthouse audits for it. Passing that audit tells you the file is present. It does not tell you whether AI tools will recommend the company.
What should a business do about llms.txt?
Writing the file takes an afternoon. Spend that time describing the site and directing buyers to pages that already address their questions. The file cannot provide answers missing from those pages, so improve them before putting more work into the guide.
The order matters because the file and the pages do different jobs. The guide directs attention to selected pages; the page text explains the business to a reader. If a page is vague, pointing an AI model toward it does not make the answer clearer. Review the pages that matter to a buyer, revise any that leave the main question unanswered, then write the guide.
My agent-readability rubric lets any firm score its own site and get the same answer. That check tells you whether a site is readable by agents. Recommendation testing asks a different question: whether AI tools name a company when asked real buying questions.
How do I know if AI tools recommend my company?
I measure recommendation rates with a locked library of a category's real buying questions. I ask ChatGPT, Claude, Perplexity, Gemini and Grok on a fixed schedule, so a later test uses the same questions as an earlier one. I check official domains so a same-named company does not count as a win. I also ask the model why it chose what it chose, and keep records in one format so week 11 compares cleanly with week 1.
During one engagement, Soapbox's AI recommendation win rate rose from 6.2% to 36.2% in 11 weeks. Trinzik delivered the work on the platform I built. This is directional evidence, not a controlled study. In another program, I published into the gaps, but the win rate did not change. I retest instead of calling it a win because results may vary between programs.
Common questions
Does llms.txt change Google Search rankings?
The available guidance does not establish what effect the file has on rankings. Before changing a page for search, check Google's guidance for the specific page information you want Google Search to understand, then make sure that information is present on the page.
How can I check whether my llms.txt file is useful?
Ask someone unfamiliar with the site to read it and identify what the company does and which pages answer buyer questions. If the guide points to pages that do not answer those questions, revise the pages before expanding the guide. This review checks whether the guide is clear to a reader.
Should I expect recommendations to rise after I publish one?
The change in my engagement is directional, and another program stayed flat after I published into gaps. Treat a change in recommendation rate as something to retest, not proof that the file caused it. Keep the same questions and schedule so you can compare the results.
Sources
If you want help checking how your company is understood on the web and by AI, contact me.