Last Updated on 6. January 2026 by Sandblue
The web is quickly shifting from “search engines that send clicks” to “answer engines that synthesize answers.” In that transition, SEOs are looking for a new control layer—something like robots.txt, but for LLMs.
That’s where llms.txt comes in: a simple Markdown file placed at the root of your domain (/llms.txt) meant to guide AI systems to your most important pages and (in theory) increase the chance of being cited in AI answers, resulting in hopefully stable or even more traffic. (Everything is better than declining. But this is reality right now.)
The big question: does it actually move traffic?
Based on our basic longitudinal test, the answer is currently: no measurable impact.
What llms.txt is (and why people care)
Think of llms.txt as a curated “map” for LLMs:
- It highlights key URLs (docs, pricing, cornerstone pages).
- It aims to reduce noise (ads, navigation clutter) and make content easier to ingest.
- It signals that your site is “AI-friendly.”
In theory, that could lead to:
- more AI crawling,
- better retrieval for RAG systems,
- more mentions and citations,
- more referral traffic.
In practice, adoption and support are still uneven—and that matters.
Our test: mid-October → December (WordPress, Switzerland)
We rolled out llms.txt in mid-October (test start), across Swiss WordPress sites totaling roughly ~7,500 daily clicks.
Result from the chart:
- Organic clicks behaved normally, no measurable increase in AIO or AI Mode via Google Search Console.
- AI referral traffic remained stable in November and December.
- We found no strong correlation between llms.txt deployment and AI referral growth.
- We did not check the impressions yet, because of the change with the removal of the num=100 parameter in September – this caused a lot of movement in the impressions in the Q4 2025.
So if you’re expecting a “step change” in AI traffic a few weeks after implementation—we didn’t see it.
Log file reality check: AI bots barely touched it
Traffic is a lagging indicator. Crawling is a leading one. So we checked server logs to see whether relevant agents even request /llms.txt.
We found no meaningful bot activity (nothing remotely comparable to /robots.txt or /sitemap.xml) from:
- Googlebot Smartphone
- Googlebot Desktop
- ChatGPT-User
- OAI-SearchBot
- PerplexityBot
- ClaudeBot
- and some more..
That’s the most important practical insight: if bots don’t fetch the file, the file can’t influence outcomes.
The “cats.txt” moment (and the illusion of control)
This is where Mark Williams-Cook now-famous “cats.txt” analogy (also mentioned at the Google Search Central Live Event 2025 in Zurich) lands: you can publish any file you want, but it only becomes a standard when major crawlers choose to support it.
That’s the core problem with llms.txt today:
- it’s community-driven,
- but crawler behavior is platform-controlled.
Implementation: easy with WordPress (Yoast/RankMath)
- Yoast feature: https://yoast.com/features/llms-txt/
- Rank Math: https://rankmath.com/kb/llms-txt/
From an ROI perspective, that’s the key:
If it’s effortless, you can enable it.
But you shouldn’t expect it to generate noticeable AI traffic—yet.
There are also multiple generators if you are not on WordPress, but you have to upload it on your client’s server.
Conclusion: Leveraging llms.txt for Competitive Advantage
llms.txt didn’t move the needle in our Oct–Dec test—no traffic lift, no observable crawler interest, no clear link to AIO / AI Mode outcomes.
Right now, llms.txt is a harmless experiment and a future-proofing checkbox, not a growth lever.
If it’s easy, enable it and move on.
If it costs engineering time, you’ll usually get better ROI from content structure, authority building, and technical cleanliness—the things AI systems already reward indirectly.
Leave a Reply