Does llms.txt do anything?
97% of llms.txt files received zero requests across 137,000 domains, and Google says the file has no effect on Search visibility. Our position, with sources, and what to do instead.
The short answer
On the available evidence, no. A May 2026 log analysis of 137,000 domains found 97% of llms.txt files received zero requests, and Google stated the same month that the file has no effect on Search visibility.
What the evidence actually says
Two findings, both from 2026, and both inconvenient for one of the most commonly sold deliverables in AI search optimisation.
The first is adoption on the reading side. A log analysis across 137,000 domains found that 97% of llms.txt files were never fetched at all. Not fetched and ignored: never requested.
The second is Google's own position, stated in May 2026, that the file has no effect on Search visibility. That covers AI Overviews and AI Mode, which are the largest AI surfaces in the market.
Our position
We publish an llms.txt on this site because it costs almost nothing to be correct, and we would rather be checkable than consistent with our own scepticism. It does not appear on our invoices, and we do not sell its implementation as a service.
Why a reasonable idea does not work yet
The proposal is sound on its face. A machine-readable manifest telling models what a site contains and how to treat it is exactly the sort of thing that ought to help, and it is modelled on robots.txt, which does work.
The difference is commitment on the other side. robots.txt works because every crawler that matters reads it and honours it, and it is a published standard in RFC 9309. llms.txt has no equivalent undertaking from any major vendor, and the request logs bear that out. A standard nothing reads is a proposal, not a standard.
It gets sold anyway, and it is easy to see why. It is a discrete deliverable, quick to produce, it demonstrates familiarity with the category, and its ineffectiveness is almost impossible for a client to detect. Those are the properties of a good line item rather than of good work.
The cost is not the file itself. It is what the file displaces. An hour spent on a manifest nothing reads is an hour not spent on the crawler directives that genuinely determine whether you appear in an answer.
What to do instead
Audit your crawler directives
Blocking a retrieval crawler removes you from assistant answers; blocking a training crawler does not. The distinction is real, measurable, and misconfigured on a large share of SaaS sites: the ChatGPT and Gemini pages set out which agent does what.
Find out which domains own your category's citations
Measurable in an afternoon on Perplexity, where the source list is visible on screen. It usually redirects the entire programme, and the citation sources programme is what acts on it.
Fix the third-party listings you control
A stale G2 profile propagates into answers on every platform. Nobody sells this because it is unglamorous, and it is routinely the largest single accuracy failure we find, as the worked audit shows.
Structure pages so an answer survives extraction
Retrieval selects passages, not pages. Answer the question in the opening passage, then qualify: that is answer engine optimisation. Real work, and nothing to do with a manifest file.
Frequently asked questions
So should we delete our llms.txt?
No. It costs nothing to keep and it is doing no harm. What it should not do is appear on an invoice, or displace work on the things that demonstrably move citations.
What if adoption changes?
Then we change our position and say so on this page, with the date. A published position that quietly moves is worse than no position.
Is this the same as robots.txt?
It borrows the idea and not the adoption. robots.txt works because every crawler that matters reads and honours it. Ours is published at surge45.com/robots.txt with a line per agent and the reasoning in the file.
About Surge45 Team
AI Search & GEO Specialists
Surge45 is the digital discovery and growth strategic advisory for SaaS. We help software companies become the answer across Google, AI search and communities, then turn discovery into pipeline. We also build WriteWorks, our content engineering platform for AI search.
Related Articles
Get More SaaS Marketing Insights
Join 10,000+ SaaS marketing leaders who receive our weekly newsletter with actionable strategies.
Subscribe to Newsletter