Does llms.txt matter? We tracked 10 sites to find out

The debate around llms.txt has become one of the most polarized topics in web optimization.

Some treat llms.txt as foundational infrastructure, while many SEO veterans dismiss it as speculative theater. Platform tools flag missing llms.txt files as site issues, yet server logs show that AI crawlers rarely request them.

Google even adopted it. Sort of. In December, the company added llms.txt files across many developer and documentation sites.

The signal seemed clear: if the company behind the sitemap standard is implementing llms.txt, it likely matters.

Except Google pulled it from its Search developer docs within 24 hours.

Google’s John Mueller said the change came from a sitewide CMS update that many content teams didn’t realize was happening. When asked why the files still exist on other Google properties, Mueller said they aren’t “findable by default because they’re not at the top-level” and “it’s safe to assume they’re there for other purposes,” not discovery.

The llms.txt research

We wanted data, not debates.

So we tracked llms.txt adoption across 10 sites in finance, B2B SaaS, ecommerce, insurance, and pet care — 90 days before implementation and 90 days after.

We measured AI crawl frequency, traffic from ChatGPT, Claude, Perplexity, and Gemini, and what else these sites changed during the same window.

The results:

• Two of the 10 sites saw AI traffic increases of 12.5% and 25%, but llms.txt wasn’t the cause.
• Eight sites saw no measurable change.
• One site declined by 19.7%.

The 2 ‘success’ stories weren’t about the file

The Neobank: 25% growth

This digital banking platform implemented llms.txt early in Q3 2025. Ninety days later, AI traffic was up 25%.

Here’s what else happened in that window:

• A PR campaign around its banking license, with coverage in major national publications.
• Product pages restructured with extractable comparison tables for interest rates, fees, and minimums.
• Twelve new FAQ pages optimized for extraction.
• A rebuilt resource center with new banking information and concepts.
• Technical SEO issues, like header structures, fixed. 

When a company gets Bloomberg coverage the same month it launches optimized content and fixes crawl errors, you can’t isolate the llms.txt as the growth driver.

The B2B SaaS platform: 12.5% growth

This workflow automation company saw traffic jump 12.5% two weeks after implementing llms.txt.

Perfect timing. Case closed. Except…

Three weeks earlier, the company published 27 downloadable AI templates covering project management frameworks, financial models, and workflow planners. Functional tools, not content marketing, drove the engagement behind the spike.

Google organic traffic to the templates rose 18% during the same period and continued climbing throughout the 90 days we measured.

Search engines and AI models surfaced the templates because they solved real problems and launched an entirely new site section — not because they were listed in an llms.txt file.

The 8 sites where nothing happened after uploading llms.txt

Eight sites saw no measurable change. One declined by 19.7%.

The decline came from an insurance site that implemented llms.txt in early September. The drop likely had nothing to do with the file.

The same pattern showed up across all traffic channels. Llms.txt neither prevented the decline nor created any advantage.

The other seven sites — ecommerce (pet supplies, home goods, fashion), B2B SaaS (HR tech, marketing analytics), finance, and pet care — all documented their best existing content in llms.txt. That included product pages, case studies, API docs, and buying guides.

Ninety days later, nothing changed. Traffic stayed flat. Crawl frequency was identical. The content was already indexed and discoverable, and the file didn’t alter that.

Sites that launched new, functional content saw gains. Sites that documented existing content saw no gains.

Why the disconnect?

No major LLM provider has officially committed to parsing llms.txt. Not OpenAI. Not Anthropic. Not Google. Not Meta.

Google’s Mueller put it plainly:

• “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.”

That’s the reality. The file exists. The advocacy exists. The adoption by platforms doesn’t show it (yet!). 

The token efficiency argument (and its limits)

The strongest case for llms.txt is about efficiency. Markdown saves time and tokens when AI agents parse documentation. Clean structure instead of complex HTML with navigation, ads, and JavaScript.

Vercel says 10% of their signups come from ChatGPT. Its llms.txt includes contextual API descriptions that help agents decide what to fetch.

This matters — but almost exclusively for developer tools and API documentation. If your audience uses AI coding assistants like Cursor or GitHub Copilot to interact with your product, token efficiency improves integration.

For ecommerce selling pet supplies, insurance explaining coverage, or B2B SaaS targeting nontechnical buyers, token efficiency doesn’t translate into traffic.

llms.txt is a sitemap, not a strategy

The most accurate comparison is a sitemap.

Sitemaps are valuable infrastructure. They help search engines discover and index content more efficiently. But no one credits traffic growth to adding a sitemap. The sitemap documents what exists; the content drives discovery.

Llms.txt works the same way. It may help AI models parse your site more efficiently if they choose to use it, but it doesn’t make your content more useful, authoritative, or likely to answer user queries.

In our analysis, the sites that grew did so because they:

• Created functional assets like downloadable templates, comparison tables, and structured data.
• Earned external visibility through press and backlinks.
• Fixed technical barriers such as crawl and indexing issues.
• Published content optimized for extraction, including FAQs and structured comparisons.

Llms.txt documented those efforts. It didn’t drive them.

What actually works

The two successful sites show what matters:

• Create functional, extractable assets. The SaaS platform built 27 downloadable templates that users could deploy immediately. AI models surfaced these because they solved real problems, not because they were listed in a markdown file.
• Structure content for extraction. The neobank rebuilt product pages with comparison tables with interest rates, fees, and account minimums. This is data AI models can pull directly into answers without interpretation.
• Fix technical barriers first. The neobank fixed crawl errors that had blocked content for months. If AI models can’t access your content, no amount of documentation helps.
• Earn external validation. Coverage from Bloomberg and other major publications drove referral traffic, branded searches, and likely influenced how AI models assess authority.
• Optimize for user intent. Both sites answered specific queries: “best project management templates” and “how do [brand] interest rates compare?” Models surface content that maps to what users are asking, not content that’s merely well documented.

None of this requires llms.txt. All of it drives results.

Should you implement an llms.txt file?

If you’re a developer tool where AI coding assistants are a primary distribution channel, then yes — token efficiency matters. Your audience is already using agents to interact with documentation.

For everyone else, treat llms.txt like a sitemap: useful infrastructure, not a growth lever.

It’s good practice to have. It won’t hurt. But the hour spent implementing llms.txt is often better spent restructuring product pages with extractable data, publishing functional assets, fixing technical SEO issues, creating FAQ content, or earning press coverage.

Those tactics have shown real ROI in AI discovery. Llms.txt hasn’t — at least not yet.

The lesson isn’t that llms.txt is bad. It’s that we’re reaching for control in a system where the rules aren’t written yet. Llms.txt offers that comfort: something concrete, actionable, and familiar, shaped like the web standards we already know.

But looking like infrastructure isn’t the same as functioning like infrastructure.

Focus on what actually works:

• Create useful content.
• Structure it for extraction.
• Make it technically accessible.
• Earn external validation.

Platforms and formats will change. The fundamentals won’t.

View the full article