Does llms.txt Work? What Google Actually Says in 2026

What is llms.txt?

llms.txt is a proposed text file, first written in 2024 and now at version 2, published at llmstxt.org. The idea is a single markdown file at the root of a domain that gives a large language model brief background on a site and links to markdown versions of its pages, similar in spirit to how robots.txt or sitemap.xml sit at the root and describe the site to automated systems.

It is a proposal, not a standard any search engine or AI company has committed to supporting. That distinction matters for everything that follows: a domain publishing an llms.txt file is opting into a format some tools might read, not a format the major platforms have confirmed they do.

Does Google Search use llms.txt?

No, and Google has said so directly. On 15 June 2026, Google Search Central clarified that llms.txt files are not needed for Google Search and will not have a negative or positive impact on a site’s visibility or rankings there. Google’s own AI optimization guide, last updated 10 July 2026, goes further and names llms.txt specifically in a mythbusting section on things site owners can ignore: llms.txt and other “special” markup or AI text files are not used by Google Search, and keeping them for other services neither helps nor hurts Google visibility.

The same guide lists a handful of other things in that ignore-list worth knowing alongside llms.txt: chunking content into tiny pieces for AI consumption, rewriting content purely to suit an AI system rather than a reader, and chasing inauthentic “mentions” of a brand. Google is explicit that unique, first-hand, genuinely useful content is what its systems weigh — not a supplementary file describing that content.

Does ChatGPT read llms.txt?

OpenAI’s own documentation does not say so. “Overview of OpenAI Crawlers” describes robots.txt controls specifically — OAI-SearchBot, which surfaces sites in ChatGPT’s search features; GPTBot, the separate training crawler; and ChatGPT-User, which visits a page on a live user request rather than to decide search inclusion. Nowhere in that documentation does OpenAI say ChatGPT search reads or uses a site’s llms.txt file to decide what to show.

Worth noting for context, not as evidence either way: OpenAI’s own developer documentation site publishes an llms.txt index of its docs. That tells you OpenAI finds the format useful for its own documentation, which is a different claim entirely from ChatGPT’s search product reading other sites’ llms.txt files to decide what to cite — and the crawler documentation makes no such claim.

So why do sites still publish one?

Reasons vary, and the proposal itself is still evolving. Some site owners treat it as a low-cost hedge: if a future tool does read llms.txt, having one already in place costs nothing to have prepared. That is a defensible reason to keep a file already built. It is a weak reason to spend real effort building one from scratch today, given that neither of the two platforms with documented crawler behavior — Google and OpenAI — has confirmed reading it for ranking or citation purposes.

Part of the appeal is also that llms.txt looks simple to produce compared with fixing indexing problems or rewriting thin pages. A single markdown file feels like a finished task in a way that a slow content or technical fix does not, and that feeling of progress is exactly what makes it easy to over-invest in relative to what it actually does. The work that is documented to matter is almost always the less satisfying kind: checking a setting, fixing a tag, writing a genuinely better page.

Where the reasoning breaks down is when llms.txt gets treated as a substitute for the things that are documented to matter: a robots.txt that actually allows the crawlers you want, pages that are indexed and snippet-eligible, and content specific and useful enough to be worth citing in the first place. A perfect llms.txt sitting in front of a thin or blocked site does nothing for either Google or ChatGPT.

robots.txt vs. llms.txt vs. sitemap.xml

These three root-level files get confused with each other because they sit in the same spot and sound similar. They do very different jobs.

robots.txt is the one file in this table with confirmed control over both platforms. If a site is choosing where to spend an afternoon on AI-visibility housekeeping, checking robots.txt against OAI-SearchBot and the other named crawlers does more documented work than building or polishing an llms.txt file.

Is it ever worth publishing one?

If it already exists, there is little reason to remove it. Google has said plainly that keeping one neither helps nor hurts Google visibility, so it is not doing damage sitting there. If it does not exist yet, building one purely to chase Google or ChatGPT visibility is spending effort against a documented non-effect on the platform that matters most and an undocumented one on the platform that matters second most.

A more honest reason to publish one: it is a small, cheap piece of housekeeping, and the AI ecosystem is still young enough that a proposal like this could pick up support from some tool later. That is speculation, not a plan, and it should be treated as such — a nice-to-have after the documented basics are handled, never a substitute for them.

The ordering matters more than the decision itself. A business that has not yet confirmed its robots.txt allows the crawlers it wants, or has pages sitting behind an accidental noindex, gains nothing by adding llms.txt first. Fix the things with a documented effect, and treat llms.txt as the last five minutes of the task, not the first.

Why we keep ours

ciaobellamarketing.com publishes an llms.txt file, and we also emit FAQPage schema on our pages. Neither is there because we believe Google ranks on it — Google has told us directly it does not, for llms.txt, and has said structured data is not required for AI search either. We keep both because they cost almost nothing to maintain and some other system, now or later, might read one of them, and there is no downside documented anywhere to having them in place.

What we do not do is let either file substitute for the actual work: robots.txt explicitly allows OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot, Google-Extended and Bingbot on our site, and the content itself is written to our house standard — answer first, question-shaped headings, a real FAQ — independent of whatever any supplementary file says about it.