Does llms.txt help Google Search? What the file actually does

That answer is narrower than “the file is useless.” The llms.txt project is a community proposal for giving tools a short Markdown map of a website. A system has to choose to fetch and use that map. Google Search has said it does not.

What the proposal contains

The proposal at llmstxt.org describes a Markdown file with a site name, a short summary, and links to useful pages. The current v2 document also supports discovery through a link relation. It remains an open proposal, not a search directive.

The file works best as a curated table of contents. A documentation site can point an agent toward the API overview, authentication guide, reference pages, and examples without asking it to interpret a large navigation system first.

It does not grant access. It does not block training. It does not replace robots.txt, a sitemap, canonical tags, or page-level controls.

Why DATIK keeps one

DATIK publishes llms.txt and an expanded llms-full.txt. The short file identifies the company, its packages, pricing, ownership terms, and current research. The longer file gives tools more context.

We keep them because the files are inexpensive to generate from the same source used for the site. They also give us a clean inventory check. When a new guide is published, its URL should appear in the hub, sitemap, and machine-readable map at the same time.

We do not report the files as a Google ranking improvement. We have no verified evidence that adding them increased a citation in another assistant either. Their value remains operational until a specific system documents support or a controlled test shows a difference.

What Google asks you to do instead

Google's guidance points back to normal search foundations. Publish useful material that adds first-hand knowledge or original evidence. Keep important pages crawlable and internally linked. Make sure the page can be indexed and shown with a snippet.

Google also rejects the idea that content must be split into tiny chunks or rewritten in a special dialect for AI. Structured data can still qualify pages for ordinary search features, but there is no special AI schema for Google Search.

If the budget only covers one task, fix the public HTML and internal links before creating llms.txt.

A maintenance checklist

Keep the file short enough to review. Link only to canonical public pages. Use descriptions that state what a page contains instead of repeating marketing language.

• the file returns HTTP 200 as plain text or Markdown;
• every listed URL resolves and remains public;
• deleted or redirected pages are removed;
• the descriptions still match the linked pages.

DATIK runs those checks with the article release because a stale inventory sends a tool toward material the business no longer stands behind.

How to evaluate a vendor claim

Ask the vendor to name the system that uses llms.txt, link to that system's documentation, and explain the measurement. “AI crawlers can read it” is not proof that a product fetches it or gives it weight.

For Google, the answer is documented. For other assistants, keep citation checks separate from Google Search Console and analytics. Record the prompt, date, model, answer, cited URLs, and whether the file was accessible.

The DATIK Content Engine maintains technical publishing assets alongside the article itself. The pricing page lists the operating fee. A free diagnostic can check whether your site needs a machine-readable map or has a more basic crawl and source problem.

Sources and review limits

• Google guide to generative AI features in Search
• Google Search documentation updates
• The llms.txt proposal, version 2

Sources were checked on September 30, 2026. This article does not claim that llms.txt improves citations, traffic, rankings, or inclusion in any assistant.