97% Of llms.txt Files Go Unread. Now What?

llms.txt just ran into its first real reality check.

On June 15, 2026, Ahrefs published a study of 137,000 sites and found that 97% of llms.txt files received zero requests in May. No bots. No humans. Nothing. That matters because llms.txt has been sold all year as a simple AI visibility win. At the same time, Google’s official AI search guidance says you do not need llms.txt files or other special markup to appear in Google Search, while Chrome’s new Lighthouse agentic browsing docs still check for llms.txt as part of machine-readiness.

That is the tension marketers need to understand.

llms.txt is not worthless. It is also not the shortcut a lot of SEO and AEO chatter made it sound like. If your goal is to show up in AI search results, citations, and recommendation prompts, the new evidence says the file is mostly hygiene and future-facing infrastructure, not a primary growth lever.

This post exists to prove one point: if your team is treating llms.txt like a magic switch for AI visibility, you are probably overinvesting in the wrong technical task.

The New Data Just Punctured A Popular AI Search Shortcut

The Ahrefs study is useful because it measures behavior, not hype.

According to Ahrefs, 28% of the 137,000 domains in its dataset publish an llms.txt file. That is a big adoption number for a format that no major AI platform has formally committed to using. But adoption and usage are not the same thing. Ahrefs found that 97% of those files got zero requests, and even among the small share that did receive traffic, 96% of requests came from bots rather than humans.

The more revealing detail is who those bots were.

Ahrefs says 77% of the bots fetching llms.txt were not AI tools at all. Many were SEO audit tools, general crawlers, tech profiling bots, or llms.txt validators. Only a tiny share of requests came from what the study classifies as AI retrieval bots. Their data also found that AI tools did not go hunting for missing llms.txt files. If the file was absent, they generally did not probe for it.

That changes the framing.

For months, many teams treated llms.txt like an AI equivalent of robots.txt, a file that major systems would naturally look for, parse, and use. The Ahrefs evidence says that assumption is weak right now. If you publish one today, the most likely outcome is that almost nobody fetches it.

That does not mean you should laugh it off. It does mean you should stop confusing potential future usefulness with current visibility impact.

Google And Chrome Are Sending Different Signals

Part of the confusion comes from mixed platform messages.

Google’s official guide to optimizing for generative AI in Search is very clear on one point: you do not need llms.txt files, AI-specific text files, or special markup to appear in Google Search, including its generative AI features. Google keeps pushing the same broader message, good AI search visibility still depends on strong SEO fundamentals, indexable pages, useful content, and eligibility for normal Search snippets.

Chrome’s docs are different because Chrome is solving a different problem.

The new Lighthouse documentation for llms.txt describes the file as an emerging convention that gives LLMs and AI agents a machine-readable summary of a site’s content. Chrome says that without it, agents may spend more time crawling a site to understand its structure and primary content. In the related agentic browsing scoring docs, Lighthouse checks for llms.txt alongside layout stability, accessibility-tree quality, and WebMCP readiness.

Those are not contradictory statements once you separate search visibility from agent usability.

Google Search is saying, “This is not required for ranking or AI inclusion in Search.”

Chrome is effectively saying, “This may still help agents navigate or interpret your site more efficiently.”

My read, based on those two sources, is that llms.txt matters more for agent workflows than for AI citation visibility today. That is an inference, but it lines up with both sets of documentation and with the Ahrefs traffic data.

Why This Matters For Agencies, Healthcare Marketers, And B2B Teams

The practical risk is not that llms.txt exists. The risk is that teams use it as a substitute for harder work.

I can already see the pattern in AI search conversations. A business hears that AI search matters, someone recommends a generator, the file gets published, and everyone acts like the technical box is checked. Meanwhile the site still has weak service pages, thin proof, muddy titles, poor internal linking, and almost no third-party authority.

That is backwards.

Ahrefs published another useful study in April showing that ChatGPT only cites about half of the URLs it retrieves, and that there is a gatekeeping layer before full page content even gets considered. Titles, snippets, URLs, and relevance to internal fan-out queries matter before a page earns citation credit. Then in May, Ahrefs tracked 1,885 pages adding schema and found no major citation uplift from the change alone.

Put those findings together and the technical lesson gets sharper.

Machines do not reward every AI-flavored artifact equally. They reward pages that are already discoverable, relevant, structurally clear, and worth citing. A machine-readable file can support that ecosystem. It does not replace it.

That is especially important in high-trust categories.

Healthcare, behavioral health, legal, finance, and B2B service businesses usually do not lose AI visibility because they forgot one file. They lose it because their authority is thin, their answers are generic, or their page structure makes extraction harder than it should be. We covered the broader operational version of that in Your IT Team Might Be Blocking AI Revenue. The same principle applies here. Infrastructure matters, but it only pays off when it supports real clarity.

At Emarketed, we have seen this on the ground. LA Roofing Materials grew from near-zero organic presence to more than 2,000 keyword rankings and a 258% increase in AI mentions through sustained SEO and AEO execution over time. That kind of result does not come from one file in the root directory. It comes from a stronger whole site.

What llms.txt Is Still Good For

The answer is not “never create one.”

The better answer is “treat it proportionally.”

If your site has documentation, API references, dense product architecture, or workflows that coding agents and browser agents may actually use, then llms.txt can still be useful as an orientation layer. Chrome’s docs point in that direction, and the Ahrefs study says coding-focused and agentic tools were more plausible readers than mainstream AI retrieval systems.

There are also three softer reasons to keep it on the roadmap.

First, it is cheap. If your CMS or platform can generate and maintain it with very little effort, the cost is low.

Second, it may matter more later than it does today. Chrome’s public push into agentic browsing and WebMCP is a sign that agent-mediated use of the web is getting more serious, even if traditional AI retrieval bots are not relying on llms.txt yet.

Third, it can force a healthy content-inventory discussion. To create a decent file, a team has to decide which pages, docs, or resources actually represent the site best. That discipline can be useful even if the file itself does not move citations.

But none of those reasons justify treating llms.txt like a front-of-roadmap AI search tactic for most marketing teams.

What To Prioritize Before You Touch llms.txt

If your goal is AI visibility in ChatGPT, Gemini, AI Overviews, Perplexity, or similar surfaces, there are more reliable priorities.

Strengthen The Pages That Machines Actually Cite

The April Ahrefs citation study found that ChatGPT’s search-sourced URLs are far more likely to earn citations than content pulled from side channels. It also showed that title relevance to fan-out queries matters meaningfully. That means your core service, comparison, FAQ, and proof pages should answer real prompts directly and quickly.

Do not bury the commercial point halfway down the page. Do not hide the specifics behind brand slogans. Do not make the title sound clever when it should sound precise.

Fix Extractability Before You Add More AI Layers

Your highest-value pages should be easy to parse. Use clear headings, strong introductory answer blocks, short paragraphs, sensible internal links, and consistent naming. If the page is hard for a worried buyer to skim, it is often hard for a machine to summarize well too.

Google’s generative AI search guidance keeps pushing marketers back to this same point: the fundamentals still carry the weight.

Build Third-Party Validation

AI systems do not learn trust from your own copy alone. They learn from the web around you. Category roundups, reviews, trade mentions, expert commentary, association profiles, and niche publisher coverage all help shape who belongs in the answer set.

If you want a broader framework for that off-site work, our AEO services sit exactly in that overlap of technical clarity, citation readiness, and authority building.

Treat llms.txt As A Final Layer, Not A First Move

Once the core pages are clean, the site is technically sound, and your off-site footprint is improving, then adding llms.txt can make sense as a support layer. If your team still wants the file, do it with light effort and realistic expectations. If you want a simple starting point, one useful option is our LLMs.txt Generator. Just do not confuse publishing the file with solving AI visibility.

The Hidden Risk Nobody Mentions Enough

There is also a security and governance angle here.

The Ahrefs study notes that researchers and bots are already probing llms.txt as a possible prompt injection surface. If agents are trained to trust these files, then a stale, sloppy, or compromised file becomes a liability, not an asset. That is one more reason to stop treating llms.txt like harmless AI confetti.

If you publish it, manage it like a controlled asset.

Version it. Limit who can edit it. Keep it factual. Link only to pages you control. Review anything a platform auto-generates. And make sure the content inside the file matches the pages your business actually wants machines and users to rely on.

For many marketing teams, that governance burden alone is enough reason to keep llms.txt modest until agent adoption is more proven.

The Smarter Position For 2026

The market does not need another fake binary here.

llms.txt is not a scam. It is also not a serious primary answer to AI visibility problems in June 2026.

The smarter position is this:

• for Google Search visibility, it is unnecessary
• for broader agent readiness, it may become useful
• for current AI citation growth, it is far down the priority list
• for technical governance, it deserves more care than the average generator pitch suggests

That is a much more useful posture than the easy headline versions.

If your site already has strong content, strong architecture, strong authority signals, and a legitimate agent use case, publish the file and monitor it. If your site is still weak where it counts, spend your time fixing the pages, titles, proof, and source footprint that AI systems already use today.

FAQ

Does llms.txt Help You Rank In Google AI Search?

Not according to Google’s current guidance. Google’s official AI search documentation says you do not need llms.txt files or other special AI markup to appear in Google Search or its generative AI features.

Should Businesses Still Publish An llms.txt File?

Sometimes. It can make sense if your site serves documentation, APIs, or workflows likely to be used by coding agents or browser agents. For most marketers chasing better AI citations, it should be a low-priority support task rather than a first move.

Why Are So Many Marketers Talking About llms.txt If It Barely Gets Read?

Because it feels like a neat technical shortcut. It is easy to generate, easy to explain, and easy to package as an AI-readiness deliverable. The Ahrefs data suggests adoption has outpaced real usage.

What Should I Fix Before Adding llms.txt?

Start with page quality, title relevance, structured page layout, service-page clarity, internal links, and third-party authority signals. Those are the factors most closely tied to whether AI systems can find, understand, and cite your content.

Is llms.txt Completely Useless For AI Agents?

No. Chrome’s Lighthouse documentation suggests it can help agents understand site structure faster, and the Ahrefs study shows some agentic and coding-oriented bots do fetch it. The point is not that it is useless. The point is that it is not a reliable AI search growth lever today.

What Should Marketing Teams Do This Week?

Audit the five pages that matter most to pipeline. Tighten the titles, improve the first 150 words, make the answers easier to extract, and check whether trusted third-party sources in your category mention you at all. If those basics are weak, llms.txt is not your bottleneck.

What To Do Monday Morning

Ask one blunt question: if we deleted llms.txt from the roadmap this quarter, what real AI visibility work would finally get attention?

For most teams, the answer will be obvious. Better service pages. Better proof. Better titles. Better off-site mentions. Better technical clarity on the pages that actually influence pipeline.

That is where the next gains are.

If you have the bandwidth, publish llms.txt as a lightweight support file and keep an eye on the logs. Just stop pretending it is the reason a brand will or will not get cited.