llms.txt: what it is and whether it matters
llms.txt is a proposed plain-text file, written in Markdown and placed at the root of your site, that lists your most useful pages for AI tools to read. It is not an official standard, Google has said it ignores llms.txt for search, and no one can show that adding one gets you cited, so treat it as optional housekeeping.
Fig.Whether AI assistants send real visitors is something to measure, not assume. Sample data, not a real site.
What is llms.txt?
The idea is simple. robots.txt tells crawlers what they may fetch. llms.txt is meant to tell language models what is worth reading. It sits at yoursite.com/llms.txt and usually contains:
- A title and a one-line description of the site.
- Short sections with links to your key pages, each with a note on what it covers.
- Optionally, links to plain-text versions of those pages.
Here is a small example for a fictional site.
# ExampleShop
> ExampleShop sells handmade ceramics. Free shipping over $50.
## Key pages
- [Shipping and returns](https://example.com/shipping): costs, times, return window
- [Care guide](https://example.com/care): how to wash and store pieces
- [Pricing and sizes](https://example.com/sizes): dimensions and prices
That fictional file is for illustration only. The proposal has been published by its author, and it is up to each AI product whether to read it.
Does Google use llms.txt?
Google representatives have said publicly that Google does not use llms.txt for search. If that holds, the file does nothing for Google Search, but check Google's current guidance because statements like this can change. Do not let anyone sell you an llms.txt as a ranking factor.
For other assistants, the picture is unclear. Some tools and developer-facing products may read the file. We have no verified evidence that ChatGPT, Perplexity, Claude or Gemini consult it when answering a question, and we will not claim they do. If a vendor tells you they do, ask for documentation.
So should you add one?
It depends on your site and, by our rough estimate, costs about twenty minutes. Our view:
- Add it if you publish documentation or developer-facing content, where a coding assistant might read your site on demand, and you can keep it current.
- Skip it if you run a small marketing site and your time is limited. Fix your pages first.
- Never rely on it for anything that matters. A file no one reads has no effect.
The downside risk is small: an out-of-date file that points to deleted pages. If you add one, put it on a reminder to review.
What matters more than llms.txt?
Everything below has a clearer link to how people and tools actually understand your site:
- Crawlable pages. Content that loads without needing a login or heavy scripts. Check your robots.txt does not block pages you want read.
- Answer-first writing. The first two sentences answer the question in the heading. See the answer engine optimization checklist.
- Facts in tables. Prices, limits and dates a reader can verify.
- Consistent facts off-site. Identical descriptions on your profiles and directories.
- Real mentions on other sites. Assistants tend to repeat what third parties say.
These steps are laid out in generative engine optimization basics for small sites.
| Source | Visitors | Revenue |
|---|---|---|
| chatgpt.com | 1,411 | $196 |
| perplexity.ai | 620 | $98 |
| claude.ai | 388 | $49 |
Fig.AI assistants as their own channel in the source table. Sample data, not a real site.
How would you know whether it helped?
This is where most llms.txt claims fall apart. You cannot isolate it. To test fairly:
- Note your AI referrals for the month before. See how to track ChatGPT referrals and the ChatGPT traffic guide, and for Perplexity the Perplexity guide.
- Add the file and change nothing else for 60 to 90 days.
- Compare visits from AI sources and your prompt log (mentioned, linked, cited).
- Accept that normal variation may be larger than any effect.
If AI visits are a handful per month, a before-and-after comparison will be noise. That is a fine reason not to bother.
You can also check your server logs for requests to /llms.txt. If nothing but your own browser ever requests it, that tells you something. Note that a normal analytics script will not record these requests, so you need access to raw logs, such as the ones your hosting provider keeps.
What are the limits?
- It is a proposal, not a standard that every tool follows.
- It can't force a citation or a recommendation.
- It describes pages. It does not make those pages good.
- Google has said it ignores it for search.
At a glance
| Question | Answer |
|---|---|
| What is it? | A proposed Markdown file at /llms.txt listing key pages |
| Official standard? | No |
| Does Google use it for search? | Google reps have said no; check current guidance |
| Evidence it earns citations? | None we can verify |
| Cost to add | Roughly 20 minutes (our estimate) plus upkeep |
| Best fit | Docs and developer sites |
| Priority versus better pages | Lower |
If you want to see how much AI traffic you get at all before deciding, Ownlytics groups AI assistant visits into their own channel, described on the AI traffic analytics page, and the live demo shows how that looks on a sample site.
FAQ
Does llms.txt improve my Google ranking?
No. Google has said it ignores llms.txt for search, so it should not be expected to affect rankings or Google's AI features.
Do ChatGPT and Perplexity read llms.txt?
We cannot confirm that they do. Some developer tools may read it, but there is no verified public evidence that major assistants use it to choose sources. Do not count on it.
What should go in an llms.txt file?
A short title and description, then links to your most useful pages with a line about each. Keep it accurate and small. Remove links to pages you delete.
Can llms.txt hurt my site?
Hardly. The main risks are stale links and the time you spend on it instead of better work. It does not expose anything that is not already public, but do not list private URLs.