Analysis
llms.txt adoption by major engines remains unconfirmed
Vendor marketing increasingly treats llms.txt as a ranking factor. No major engine provider has publicly confirmed using it as a retrieval input.
llms.txt now appears routinely in AEO vendor marketing, frequently framed as a lever for improving AI visibility. The evidence for that framing has not kept pace.
What is established: the format has real traction among AI-native developer tools and documentation platforms, several of which clearly consume it. It is trivial to produce, costs nothing to serve, and the exercise of writing one — deciding which twenty pages matter and describing each in a line — is a useful content audit in itself.
What is not established: that any major consumer assistant uses it as a retrieval input. We are not aware of a public confirmation from OpenAI, Anthropic, Google or Perplexity that llms.txt influences what their systems retrieve or cite. Absent that, claims that publishing one will raise your citation rate go beyond the available evidence.
Two clarifications worth repeating, because both recur in practitioner discussion:
It is not a standard. There is no specification body, no conformance test, and nothing you can be compliant with. It is a community proposal that spread by adoption.
It is not `robots.txt`. robots.txt withholds or grants crawl permission and is broadly respected. llms.txt offers a curated content guide and carries no access-control meaning at all. Blocking GPTBot in robots.txt while listing the same pages in llms.txt accomplishes precisely nothing — a configuration we see described as a fix more often than it should be.
Our position: publish one, spend an afternoon on it, keep it generated rather than hand-maintained so it does not rot, and then go and check your crawler access, which is where measurable gains actually come from.
Full treatment in llms.txt Explained. This site's own llms.txt is generated from published content on every request, so it cannot drift out of date.