A data verdict
Is llms.txt worth it?
Short answer: add it if you like — but do not expect it to get you cited. The idea is appealing: a plain-text file that tells AI models what your site is about. The measured reality is that its effect on whether assistants name you is unproven, and the signals that actually decide citations are elsewhere. Here is the honest picture, with the numbers.
What llms.txt is, in one paragraph
llms.txt is a proposed convention — a Markdown file at the root of your
site that lists your key pages and a short description, meant to give language models a
clean, curated view of your content. It is a reasonable idea. The question is not whether
it is tidy; it is whether it changes what an AI says about you. On that, the
evidence is thin.
The verdict: no proven effect on citations
Three things, taken together, are why we don't sell llms.txt as a lever:
| What was checked | What it found |
|---|---|
| Correlation study, ~300,000 domains (SE Ranking) | No link between having llms.txt and being cited in AI answers. |
| Crawl-log analysis of AI bots (Ahrefs) | The file is almost never fetched — reported around 97% of llms.txt files never requested by an AI crawler. |
| Official support from the big assistants | None announced. No major vendor states its crawlers use llms.txt as a citation or ranking signal. |
A file that the bots don't read, that no vendor commits to, and that shows no measured correlation with citations is not a growth lever. It is, at best, housekeeping.
These figures come from published third-party analyses (SE Ranking, Ahrefs), not from our own measurements; studies can be redone and a vendor could add support tomorrow. We'll change this verdict the day the evidence does. What we measure ourselves is the part below — which sources answers are actually built from.
Why the appeal doesn't survive contact with the data
The pitch for llms.txt assumes models reach for a special file to learn about you. In practice, when an assistant recommends a company it is retelling a source it read — and those sources are maps, directories and ordinary web pages, not a root-level text file. In our own measurements, a brand's own site appears among the cited sources in only about a third of answers; the rest lean on listing platforms. A file almost no crawler requests sits even further from the answer than your homepage does.
What actually moves citations
If the goal is to be named, the levers are measured and unglamorous:
- Classic search position. Outside the top of ordinary results, engines rarely pull a source into the answer — search visibility is the entrance, not a replacement.
- Presence in the sources the engine reads. For most niches that means the maps and directories an engine cites; which ones differ by field and by engine, and that is measurable. See which sources engines cite.
- A clear category description. Being described in the words a buyer uses — by category, not just by name — is what lets the model match you to the question rather than skip you.
So should you add it? Yes — and expect nothing
It is cheap, standard and harmless. A correct llms.txt won't hurt, and we publish one ourselves. Two honest rules:
- Add it, don't invest in it. Spend an hour, not a budget — and never pay for llms.txt "optimization" as a citation service.
- Don't credit it for wins. If your AI visibility improves after you ship one, the cause is almost certainly the search and directory work you did alongside it, not the file.
Where a machine-readable layer does earn its place is as a data asset: structured, open figures (we publish ours as a dataset and JSON feed) that others cite and link to. That builds the kind of signal engines reward — references — rather than betting on a file they don't read.
Measure what really cites you
Free, no sign-up: your buyers' questions across several engines, the platforms their answers are built from, and whether your brand appears in those sources.
Run a free checkCommon questions
- Does llms.txt help my brand get cited by AI?
- No measured evidence that it does. A ~300,000-domain study found no correlation with citations, crawl analyses report around 97% of files never fetched by AI bots, and no major assistant has announced it uses the file as a signal. Treat it as unproven.
- Should I add llms.txt anyway?
- Yes — it is cheap and harmless, and we publish one. Just don't expect it to change whether AI names you, and don't pay for llms.txt "optimization". Put the time into search position, directory presence and a clear category description.
- What actually makes AI assistants cite a brand?
- Position in ordinary search, presence in the maps and directories an engine reads for your niche, and a clear description against the category. A machine-readable data layer helps as a citable asset and link magnet, not as a trick.