Google ignores your llms.txt
Google's documentation says Search does not use llms.txt, and that having one neither helps nor harms you. That is worth knowing before you ship one, and worth understanding after.
There is a file a lot of sites added over the last two years, on the theory that AI search would want it. Google has now said, in its own documentation, that Search does not read it.
The wording is not ambiguous. Google Search does not use llms.txt files; creating one will neither harm nor help your visibility or rankings in Search; and if you keep one for some other service, that is fine. The AI features documentation puts the general version of it in one sentence: “You don't need to create new machine readable files, AI text files, or markup to appear in these features.” It adds that there is no special schema.org structured data to add either.
Why the file spread anyway
llms.txt is a reasonable-looking idea. It borrows the shape of robots.txt, which is a real
standard that real crawlers obey, and applies it to a new audience. It costs an afternoon.
And unlike almost everything else in this field, shipping it feels like finishing something.
That combination is the whole mechanism. A cheap artifact that resembles a standard, aimed at a system nobody can inspect, is very hard to argue against — because the argument requires evidence about a black box, and the file requires none.
It also arrived at a moment when practitioners badly wanted a lever. Nobody knows how to influence an AI answer the way they know how to influence a blue link, so a file you can create and then point at fills a real psychological gap. It just does not fill a technical one, at least not at Google.
What actually decides whether AI features quote you
The same page that dismisses the file states the requirement in one line: to be eligible for AI Overviews and AI Mode, “a page must be indexed and eligible to be shown in Google Search with a snippet, fulfilling the Search technical requirements.”
Read that as a dependency chain, because it is one. Eligible for a snippet depends on being indexed. Being indexed depends on being crawled and then kept, which is a decision Google makes about your content and the place most sites are actually losing. No file placed at the root of your domain moves any of those.
Which means the levers you already have are the levers. If you want to control how your
content appears in AI features, the documented controls are nosnippet, data-nosnippet,
max-snippet and noindex — the same ones that governed snippets before any of this — plus
the generative AI control in Search Console if you want out of those surfaces entirely.
So should you delete it?
Not necessarily, and this is where honesty costs us something: we ship a tool that checks llms.txt files, and we are keeping it. Google is not the only consumer of a site. Some documentation platforms and some agent frameworks do read these files, and if one of yours does, the file is doing a job.
What changes is where it sits in your list. It is an integration detail for specific services you can name, not an SEO measure. If you cannot name the service that reads yours, you have a file nobody reads, and the twenty minutes a quarter you spend maintaining it are twenty minutes that could go to the indexing problem instead.
The test worth keeping
This will happen again. The next well-shaped file, the next markup vocabulary, the next required-by-AI convention will arrive with the same properties: cheap, plausible, aimed at something opaque.
Two questions handle almost all of them. Who reads this, by name? And what does the platform say about it in its own documentation? If the answer to the first is “AI” and the answer to the second is nothing, you are looking at a habit forming, not a technique.
Frequently asked questions
Does having an llms.txt hurt my rankings?
No. Google's guidance is explicit that these files neither harm nor help visibility or rankings in Search, because Search ignores them. It is a neutral file, not a risky one.
Do other AI systems use llms.txt?
Some tools and platforms do read it, which is why Google's guidance says keeping one for other services is fine. The useful discipline is to be able to name which service you are maintaining it for. “AI” is not a name.
What about serving Markdown versions of my pages?
Same answer from Google: you do not need to create Markdown or other machine-readable files to appear in Search, including its generative features. It can still be a good idea for other reasons — this site serves a Markdown copy of every post because agents read them more cheaply than HTML — but it is a developer-experience decision, not a ranking one.
How do I stop AI Overviews from using my content?
With the controls that already existed: nosnippet, data-nosnippet, max-snippet or
noindex, depending on how much you want to withhold. There is also a generative AI control
in Search Console that excludes your site from those features; Google states it is not used
as a ranking or inclusion signal elsewhere in Search.
Is Google-Extended the same thing?
No. Google-Extended governs whether your content is used for training in other Google AI products. It is a separate control from whether AI Overviews and AI Mode can quote an indexed page, and it is not an SEO setting.
Sources
- AI features and your website — no machine-readable files needed, the snippet-eligibility requirement, and the available controls
- Google's guide to optimizing for generative AI features — “optimizing for generative AI search is optimizing for the search experience, and thus still SEO”
- Search generative AI control — opting a site out, and what it does not affect
- Latest documentation updates — the June 15, 2026 entry clarifying guidance on llms.txt files