Technical
llms.txt Generator
The file, and an honest account of what it does
Point it at your site and get a formatted llms.txt: a project summary, sections of titled links with one-line notes, and the plain statement that Google has said this file has no effect on Google Search, AI Overviews or AI Mode.
Get early accessWhat it returns
- A formatted llms.txt: an H1 with the project name, a short summary, then sections of markdown links each carrying a one-line description.
- A candidate list of the pages worth including, drawn from your own structure, rather than a dump of every URL the site has.
- The plain statement that Google has said llms.txt has no effect on Google Search, AI Overviews or AI Mode, kept next to the output rather than buried.
- Which crawlers and products are actually known to request the file, so publishing it is a decision made against a real audience.
When this matters
Someone forwards a thread claiming llms.txt is how you get into AI answers, and you are asked for one by Friday. The honest position is narrower than the thread: Google has said publicly that the file has no effect on Google Search, AI Overviews or AI Mode, so shipping it will not move anything there. It is a small static file that some other tools and agents will read, and the cost of publishing it is close to nothing. Worth doing with the right expectation, not as a ranking tactic.
How to answer this today
AI visibility is the measurement side, and it answers the question actually behind the request: it runs your prompts against several answer engines, samples each one more than once because the engines are not deterministic, and records which domains were cited. If publishing a file changed anything, that is where it would show up, which makes it equally the place to find out it did not.
AI visibilityWhy this is harder than it looks
Google has said it does nothing
Google has stated that llms.txt has no effect on Google Search, AI Overviews or AI Mode. Any page implying otherwise is selling something. The file can still be useful to other consumers, which is a much smaller and much more defensible claim.
It is not a standard
No search engine or model provider requires llms.txt, there is no validator anything has to pass, and the format is a proposal that implementations follow loosely. The output is a convention, not a specification.
A list of links is not a summary
Whatever value the file has sits in the one-line descriptions that tell a reader which page answers what. Those cannot be derived from a sitemap, which is why most published examples are close to useless.
It drifts immediately
The file is static and the site is not. A hand-written llms.txt is accurate the day it ships and quietly wrong two quarters later, pointing at URLs that have moved. Anything worth publishing has to be regenerated rather than written once.