What Google Actually Says About llms.txt
Vorec Team · 2026-09-27 · About 6 min read
Your docs site has an `llms.txt` file. Someone added it after reading that AI search needs one. It is tidy, it is correctly formatted, and as far as Google Search is concerned it is doing nothing for your visibility.
That is not a guess. Google states it plainly in its own documentation, and much of the advice recommending the file does not mention that statement.
This post is what the primary sources actually say. Checked on 27 September 2026, with links so you can read them yourself rather than take our summary of them.
What is llms.txt?
`llms.txt` is a proposed standard: a Markdown file at the root of a domain listing key pages with short descriptions, intended to help large language models find and understand a site's content. Think of it as a curated index written to be read rather than parsed.
The idea is reasonable. Documentation sites sprawl, and a short list saying "here are the pages that matter and what each is for" is useful to anyone trying to understand a product quickly.
The question is not whether the idea is sensible. It is whether anything you care about actually consumes the file.
What Google's documentation says
Google's AI features and your website guidance, last updated 10 July 2026, addresses this directly. Paraphrasing its three relevant points, with the link above so you can check the wording yourself:
- You do not need to create special AI-specific files, AI text files, markup or Markdown to appear in Google Search, including its generative AI features, because Search does not use them. This is about files created for AI systems — it is not a statement that Google ignores ordinary structured data such as schema.org markup, which Search does use.
- Creating such a file neither helps nor harms your visibility or rankings in Google Search.
- The same page also says there is no ideal page length, and no requirement to break content into small pieces for AI to understand it.
We paraphrase rather than block-quote deliberately: a paraphrase we can stand behind, with the source linked and the date we checked it, is more useful than a quotation you would have to verify anyway. Read the page.
The scope here is Google Search specifically. Google saying it does not give these files special treatment is not the same as saying no system anywhere fetches them, and Google notes such files can still be discovered and crawled like any other URL. "Nothing reads it" would be a stronger claim than the evidence supports.
So should you delete your llms.txt?
Not necessarily — but be honest about why you are keeping it.
- Do not create it expecting Google Search or AI Overviews visibility. The primary source says that will not happen.
- Do not let it displace real work. A curated index of pages that are not indexed is worth less than fixing the indexing.
- Do keep it if it is genuinely useful — as a human-readable map of your docs, or if a specific tool you rely on consumes it.
We keep ours. We do not count it as an SEO measure and we do not report it as one.
The part people skip
The same Google guidance is clear that the mechanism has not changed: crawlable, indexable, genuinely useful content is what surfaces, including in generative features. That points at less exciting work.
| Instead of | Do this |
|---|---|
| Adding `llms.txt` and waiting | Check whether your key pages are indexed at all |
| Chopping prose into extractable blocks | Answer the page's question clearly, near the top |
| Writing "for AI" | Write for the person with the problem |
| Assuming a file grants visibility | Confirm crawlers receive your actual content |
One clarification on row two, because it is easy to over-read: Google saying there is no requirement to chunk content is not an argument against headings, structure or short paragraphs. Those help human readers, which is the reason to use them.
On citing studies, including the one everyone cites
If you read about AI visibility you will meet Ahrefs' study of AI Overview brand visibility, usually via its headline that brand mentions correlate far more strongly than backlinks.
The headline is real. The sample it came from was selected like this:
- The sample was filtered to domains with Domain Rating above 40
- It also required a top keyword with at least 800 monthly searches
- It reports analysing the roughly 74% of brands that had non-zero AI Overview mentions, having excluded the ~26% with none
- The authors state correlation is not causation, and describe the factors measured as showing moderate to very weak correlation
- It is about Google AI Overviews, not AI platforms generally
None of that makes the study wrong. It describes domains above DR 40 with a top keyword at 800+ monthly searches — that is the population, stated in the study's own terms rather than our guesses about company size or site age. Whether the findings transfer to sites outside those thresholds is a question the study does not address. Treating that as a limit on applicability is our reading, not theirs.
The general habit is the useful bit: when a post cites a study, check the inclusion criteria before applying the conclusion to your own site.
How to check the claims here yourself
- ] Open [Google's AI optimization guidance and read the section on machine-readable files
- [ ] Note the "Last updated" date — guidance changes, and a post from six months ago may describe an older version
- [ ] Check your own indexation in Search Console before optimising anything else
- [ ] For any study you are about to act on, find its sample selection section
The wider pattern
Advice hardens into received wisdom quickly. Someone proposes a standard, a post speculates it might matter, a second post reports the speculation as a recommendation, and a few repetitions later it is "best practice" — sometimes while the primary documentation says something different.
The defence is dull but reliable: read the vendor's own docs, link them, note the date, and mark clearly which parts are your inference rather than theirs.
FAQ
Does llms.txt help with ChatGPT or Perplexity?
We read OpenAI's crawler documentation and Perplexity's crawler documentation on 27 September 2026 and found no statement either way. These checked pages alone do not establish whether `llms.txt` benefits those systems — in either direction.
Will adding llms.txt hurt my site?
Per Google's documentation, no. The real cost is opportunity cost.
Do I still need a sitemap.xml?
Google documents sitemaps as a supported discovery mechanism, and says they are not required for every site — small, well-linked sites often do fine without one. The relevant point here is narrower: `llms.txt` is not a substitute for a sitemap, because Google documents using one and not the other.
Does this mean I should avoid JavaScript?
No, and that is a separate question. Google supports JavaScript and documents crawling, rendering and indexing as distinct stages, with rendering deferred until resources allow. Making sure content is available without requiring client-side rendering is a compatibility decision with trade-offs of its own, not something Google forbids.
Related reading: Do AI Crawlers Run JavaScript?.
Want the recording to narrate itself? Record with Vorec — it has its own macOS recorder, and an AI agent can drive it for you — or upload a recording you already have. Vorec drafts narration matched to the workflow it captured and generates the voiceover, so nothing is spoken into a microphone. The same capture can also produce a written step-by-step guide. Start free — 7-day trial, 100 credits, no credit card required. Trial includes up to 3 projects; exports carry a watermark.