Does Your Local Business Need an llms.txt File?

What an llms.txt file actually does

Don't get hung up on the idea of llms.txt: it's a good idea, and well-intentioned. But it's not a rule.

The llms.txt convention was proposed as a way to help language models and autonomous agents have a clean, very compact website map. This file lives in the root directory of a domain, just like robots.txt, but with a different purpose.

A typical file contains a short description and links to important pages in Markdown, the plain-text format often used for documentation. A developer tool or browsing agent could use that map instead of working out the site's structure from scratch. The proposal is useful context, but it is still a proposal. It does not require an AI company to participate.

It's not a permission, nor a gate, as robots.txt is for crawler access. The llms.txt file is not similar to a sitemap submission: there is no promise or guarantee it will be seen, read, or used by AI search tools to cite or recommend your business. Think of the llms.txt file as a tourism brochure a business owner may put on a counter: it's optional, and there's no law that an incoming visitor must read it. It's a convention, not a standard.

That distinction matters when a sales report describes the file as a technical requirement. A valid llms.txt file can exist while important service pages are blocked, outdated or vague. It cannot repair those pages, submit them for indexing or tell a customer whether the business is a good fit.

What the major platforms publicly say

Current site-owner guidance draws a fairly clear line between documented controls and an optional file.

Some AI companies publish llms.txt files for their own developer documentation. That shows the format can be a useful compact directory. It does not show that the same companies' search and recommendation products look for the file on an electrician's, dentist's or restaurant's website. Publishing a file and promising to consume other sites' files are different acts.

Public documentation cannot expose every private product decision, and it can change. The careful conclusion is not that no tool will ever read llms.txt. It is that the major products do not currently give a local business a documented visibility reason to prioritize it.

• Google: Google Search says it does not use llms.txt, including for generative AI features. Google also says the file will neither help nor harm Search visibility or rankings.
• OpenAI: OpenAI tells publishers who want their content eligible for ChatGPT Search to allow OAI-SearchBot in robots.txt. Its publisher guidance does not offer an llms.txt setting or promise a benefit from publishing the file.
• Anthropic: Anthropic documents ClaudeBot and Claude-User access through robots.txt. Its site-owner guidance does not say an llms.txt file improves a business's chance of appearing in Claude.

The traffic evidence is weak too

Ahrefs checked 137,210 domains using its analytics products and found valid llms.txt files on roughly 38,000 of them. Ninety-seven percent of those files received no request during the month studied. The sample leans toward technically engaged site owners, so it should not be treated as a census of the web.

The small share of requested files needs careful reading too. Many requests came from audit tools, general crawlers and other systems rather than search assistants. Note that the mere fact that a request was received doesn't mean that the file would have any impact on the results of any search. A server log can show that a URL was fetched. It cannot show that the file changed an answer or caused a recommendation.

Put the published guidance and the traffic study together and the owner decision becomes simpler. There is a plausible use for a compact machine-readable directory, especially for large documentation sites and coding agents. There is not yet a sound basis for treating it as required local SEO work.

When the file may be worth adding

Consider adding the file only if it meets one or more of the following relatively narrow conditions:

Even then, someone has to own it. A generator may make the first version free, but a stale file creates another public description of the business that can disagree with the live site. Remove services that have ended, old location pages, retired offers and unsupported claims. Do not put private documents or instructions in a public file.

A local business with ten stable pages may gain almost nothing from another map of those pages. A software company with thousands of technical documents has a more credible use. The value depends on the site and the reader, not the fact that the filename mentions language models.

• You are on a hosting platform or CMS that automatically creates and updates the file at low cost, and you don't mind letting it do its thing.
• You are operating a large documentation or support site, and you're interested in a condensed map for use by developers and code-based agents.
• You can inspect server logs and want to run a bounded experiment without displacing a more useful job.

Check the file before paying for it

If an agency includes llms.txt in a proposal, ask four plain questions: Which product is expected to read it? What official documentation supports that expectation? What observable result will be measured? Who updates the file when the business changes? A promise about rankings or recommendations should come with stronger evidence than a screenshot of a successful file check.

You can inspect an existing file without a special tool:

A missing file is not an error unless a system the business actually depends on requires one. A clean automated file can stay. A broken or misleading one should be corrected or removed.

1. Add `/llms.txt` to the main website address and open it directly.
2. Confirm that it returns a real text or Markdown file, not a branded HTML error page.
3. Compare its business name, services, service area and links with the live website.
4. Remove obsolete links, duplicate material, private information and claims the site cannot support.
5. Check server logs for requests if that access is available. Treat a fetch as a fetch, not proof of influence.

Spend the next hour on the website itself

An ordinary local business has better uses for the next technical hour. Work down the customer path before adding an optional AI file.

The website access guide explains how to check ChatGPT's search crawler without confusing access with preference. The business-facts guide covers the details that should agree across the site and major profiles. Both jobs can expose a customer-facing problem. An llms.txt file usually cannot.

Add the file if the site can generate an accurate version cheaply, someone will maintain it and no more urgent problem is waiting. Otherwise, defer it. Fix the dead booking link, unclear service area or unsupported service claim first.

• Make sure the homepage and important service pages load, remain crawlable and use stable canonical URLs.
• State the real services, service area and important limits in words a customer understands.
• Keep the phone number, hours, address, booking path and major business profiles accurate.
• Publish first-hand proof that helps a buyer judge the work, such as a documented project, named qualification or specific service process.
• Maintain ordinary internal links and the XML sitemap so important pages are discoverable.