LLMs.txt is a proposed plain-text file placed at a site's root that lists the pages you want language models to read first, and as of mid-2026 it has no measurable effect on how AI search engines rank, cite, or include your content. Some crawlers fetch it. None of the major answer engines have confirmed they weight it. Google has stated that standard SEO signals, not a dedicated LLM file, drive inclusion in AI Overviews [S1]. So the honest position for a content team is this: llms.txt is a low-cost, forward-looking bet, not a ranking lever you can pull this quarter.

That gap between the promise and the observed behavior is the whole story. Below is what the file is, why it currently does nothing for GEO, and how to decide whether it is worth the ten minutes.

What is llms.txt?

llms.txt is a Markdown-formatted file, served at `yourdomain.com/llms.txt`, that acts as a curated map of your most valuable content for language models. Think of it as a sitemap written for reading comprehension rather than crawl discovery. A typical file opens with an H1 site name, a short summary, then grouped links to documentation, FAQs, how-to guides, and reference material, each with a one-line description.

The format was proposed in late 2024 by Jeremy Howard of Answer.AI as a way to help models find clean, high-signal pages without wading through navigation, ads, and boilerplate [S2]. Two variants exist. The base `llms.txt` lists links. An optional `llms-full.txt` inlines the actual page content into one document so a model can ingest everything in a single fetch.

The intent is reasonable. The problem is adoption on the consumer side. Publishing a file is one thing. Getting ChatGPT, Perplexity, Gemini, and Google's AI systems to read and act on it is another, and that second half has not happened at any scale worth planning around.

Why does llms.txt have no impact on AI search today?

Because none of the systems that decide AI search visibility have committed to reading it, and log data does not show preferential treatment where the file exists. Requests for llms.txt do appear in server logs across large property networks, but they are handled like any other text asset: fetched, sometimes indexed, and given no observable ranking or citation advantage tied to their presence [S3].

There are three practical reasons the file stays inert.

First, the major answer engines already run their own retrieval pipelines. ChatGPT search, Perplexity, and Google's AI surfaces build context from live crawls, existing search indexes, and their own ranking models. They do not need a publisher-authored shortcut, and trusting one would open an obvious manipulation vector. A file you write about your own content is a claim, not a verified signal.

Second, there is no enforcement or validation layer. robots.txt works because crawlers agreed decades ago to honor it and there are consequences for ignoring it. llms.txt has no such contract. A model can read it, misread it, or skip it entirely with zero downstream effect.

Third, the economics do not favor it yet. Building support means engineering, testing, and abuse handling for a standard that a minority of sites have adopted. Google reported that its AI Overviews already reach over a billion users, drawing from the same web index that powers classic search [S4]. With that pipeline working, there is little pressure to bolt on a parallel, spoofable input.

So GEO in 2026 still rests on the fundamentals: crawlable pages, clean structure, direct answers near the top, factual claims a model can lift with confidence, and enough authority that engines trust the source. A pre-publish gate like Dokeo checks those exact properties, whether or not an llms.txt file exists on the domain.

When does llms.txt actually help?

It helps in the one place it was genuinely designed for: agent and RAG workflows that you or a partner control. If a browsing agent, an internal documentation assistant, or a custom retrieval pipeline is pointed at your site, an llms.txt file gives it a clean starting index and cuts the noise of parsing full HTML. That is a real, present-day benefit, just not an AI-search-ranking one.

Three situations make the file worth publishing now:

Developer and API documentation. Coding assistants and doc-aware agents benefit most from a structured list of canonical reference pages. This is where early adoption clusters, and where the payoff is clearest.
Large content libraries. If you run hundreds of articles, a hand-picked list of your best twenty or thirty pages tells any cooperating system what to prioritize.
Partner or internal RAG. When you know the consumer respects the file, you get deterministic behavior instead of hoping a crawler guesses right.

Outside those cases, treat the file as a cheap hedge. It does no harm when maintained carefully, and if a major engine later announces support, you are already positioned. Adoption takes minutes. That is the correct frame: insurance, not investment.

How should content teams treat llms.txt right now?

Publish it if it is cheap to maintain, but do not reallocate a single hour from work that actually moves AI visibility. The ranked priorities are unchanged.

Start with content that answers questions directly and early, because answer engines extract and cite the first clean, complete statement they find. Structure pages with real headings, including question-form headings that match how people prompt. Attach clear sourcing to every factual claim so a model can quote you without hedging. Keep pages fast and crawlable so the systems that do drive citations can reach them. Then, and only then, drop an llms.txt at the root and move on.

One maintenance warning. A stale llms.txt is worse than none. If it points to dead URLs or outdated pages and a future engine starts honoring it, you will have hard-coded bad signals into your own site. Put it on the same review cadence as your sitemap, or leave it off.

Frequently asked questions

Does llms.txt help my site rank in ChatGPT or Perplexity? No, not today. Neither ChatGPT nor Perplexity has confirmed it reads llms.txt as a ranking or citation signal, and log analysis shows no advantage tied to the file's presence [S3]. Your visibility in those tools still depends on standard crawlability, answer quality, and source authority.

Is llms.txt the same as robots.txt? No. robots.txt tells crawlers what they may not access and is broadly honored. llms.txt is a suggested reading list for models, with no enforcement and no committed consumers among the major engines. They solve different problems and one does not replace the other.

Should I add llms.txt anyway? Yes, if maintaining it costs you almost nothing, especially for documentation sites and agent-facing content. Keep it current, point it only at pages you would stand behind, and expect zero AI-search lift from it in the near term.

Sources

[S1] Search Engine Land, "Google: llms.txt is not used for AI Overviews," https://searchengineland.com/google-llms-txt-ai-overviews
[S2] Answer.AI, "The /llms.txt proposal," https://llmstxt.org
[S3] Promptwatch, "What our crawler logs show about llms.txt," https://promptwatch.io/blog/llms-txt-logs
[S4] Google, "Generative AI in Search: reach and rollout," https://blog.google/products/search/generative-ai-overviews-reach

LLMs.txt and GEO: Why the File Is Ignored by AI Search Today

Human or LLM