Skip to content
FlatRelay

llms.txt: an optional experiment, with limits.

By FlatRelay · Updated

Practical guidance from a consultancy selling SEO and AI visibility services. Sources, examples, and limitations are identified in the text. Editorial standards and corrections.

In short

llms.txt is a plain-text or Markdown file at your site root (/llms.txt) that gives AI systems a clean, curated map of your business and your most important pages. It is an optional convention; Google does not require it for Search, AI Overviews, or AI Mode.

What it can and cannot do

A curated text file can document your site for tools that choose to read it. It cannot replace indexable HTML, useful content, internal links, or verified business information. We do not treat its presence as a ranking factor or promise that an AI engine will use it.

Google’s guidance on AI features says no special AI files or markup are required. Treat llms.txt as an experiment and measure any claimed benefit separately.

What to put in it

  • A one-paragraph description of the business (what you make or supply, for whom, where).
  • Links to your most important pages with a short note on each (key products and specs, certifications, how to request a quote).
  • Optional sections for documentation, FAQs, and policies.

Mini structure

Illustrative content, not a real supplier profile. Replace all claims with verified facts. Use absolute canonical links in a production file so readers and tools can resolve destinations independently.

# Acme Extrusions
> ISO 9001 aluminium extrusion manufacturer; low- to mid-volume; ships EU & SEA.

## Key pages
- /products/extrusions: profiles, alloys, tolerances, MOQs
- /certifications: ISO 9001, RoHS, REACH
- /quote: how to request an RFQ (specs needed, typical lead times)

Want to see a real one? This site's own llms.txt is built exactly the way we recommend.

When the file is worth maintaining

Consider it when a known tool or customer workflow actually consumes a curated documentation index. Start from that use case: which consumer reads the file, which pages it needs, and how you will detect stale links. If no consumer is known, keep the experiment small and prioritize useful public pages first.

The llms.txt project describes the proposal and its evolving format. Its existence does not establish adoption by every search engine. Check the intended consumer's documentation before relying on format-specific behavior. A file that can be fetched is not proof that an engine uses it in an answer.

Practical maintenance checks

  1. Open the public URL and confirm it returns readable text successfully.
  2. Follow every listed link and replace redirects with the canonical destination where practical.
  3. Compare the business summary, prices, and capabilities with the public source pages.
  4. Remove private documents, credentials, and information that is not intended for public access.
  5. Assign a person or release check to update the index when the source pages change.

The file is public documentation, not an access-control mechanism. It cannot override robots.txt, authentication, or a firewall rule. Publishing a link does not grant permission to expose confidential engineering drawings or customer records.

Measure the claim you are making

Separate three questions: can the file be retrieved, did a known consumer request it, and did a useful outcome change? An HTTP response answers only the first. Server logs may support the second, subject to bot verification. A citation change requires preserved observations and a comparison that accounts for other site edits.

If you publish the file and rewrite several product pages together, a later improvement cannot be attributed to the file alone. Use the visibility measurement protocol to disclose test conditions. Keep null results, and do not turn a successful retrieval into a claim about rankings.

Common questions

Is llms.txt an official standard?

It is a proposed convention, not a Google Search requirement. Support varies by system, and its presence is not evidence of better rankings or citations.

Does it replace good HTML and schema?

No. It complements them. The page content and structured data still do the heavy lifting.

All resources

See where your site stands today.

Get a free audit, or the full report and Score.