Does Your Website Need an llms.txt File?

By Ryan Goering, founder and CEO of BaaDigi, a contractor marketing agency.

Last updated: September 2026. We are evaluating a proposed convention here, not claiming that our clients gain citations simply by adopting it.

What does the current proposal specify?

The llms.txt v2 proposal, updated August 2026, supports a file at the root or a subpath such as /docs/llms.txt. Its only required section is an H1 naming the site or project. It describes optional summaries and sections linking to supporting material, and recommends links to agent-friendly content such as Markdown versions.

That distinction matters when reviewing old advice. “It must be at the root or it cannot be found” is too broad for the current proposal. Likewise, a particular checker’s rules should not be presented as the proposal itself. Ask which specification and tool version a vendor is using.

How is it different from other website files?

Keep the business purpose of each mechanism clear. Adding one does not repair the absence or misuse of another.

For crawler preferences, use the separate AI crawler controls guide. This article does not recommend changing training preferences merely because you decide to publish a resource overview.

What would a useful business example look like?

The following is an illustrative structure using example.com, not a claim about a real contractor. Replace it only with accurate descriptions and working public destinations that you own or have selected deliberately.

# Example Service Business

> Public information about services and contacting the team.

## Services
- [Service overview](https://example.com/services): Work the business offers.
- [Service area](https://example.com/service-area): Places the team actually serves.

## Contact
- [Contact details](https://example.com/contact): Current ways to reach the office.

Do not copy the example as evidence that your own file is complete. The editorial question is whether each link gives an agent useful, accurate information. If your CMS offers maintained Markdown versions, evaluate those destinations. Do not invent alternative URLs that the server does not provide.

Keep promotional language out of the resource descriptions. A label explaining the page’s purpose is more useful than an unsupported claim that the business is the best in its region.

What does Lighthouse currently check?

Chrome’s current Lighthouse documentation says the audit flags server errors when retrieving llms.txt and treats a missing file returning 404 as not applicable. That documented check does not establish that a file has accurate prices, safe links or comprehensive business information.

Keep technical and editorial checks separate. A tool may successfully retrieve a file that links to obsolete pages. A file may be absent without indicating a broken customer journey. Read the actual result instead of turning every missing item into a sales emergency.

Ask for the test name, version, tested URL and output when someone reports a failure. If another tool uses additional validation, describe those checks as that tool’s behavior. Do not generalize one implementation into an industry requirement.

How should you review a file before publication?

Assign an owner and test the actual destinations. Our recommended review focuses on usefulness and maintenance rather than satisfying a decorative score.

1. Confirm that the name and summary accurately describe the business.
2. Choose a small set of resources that answer meaningful questions.
3. Open every destination and remove broken, obsolete or unintended links.
4. Check that no private customer or operational material is exposed.
5. Confirm the live file is served as intended rather than a branded error page.
6. Record who updates it when services or URLs change.

If the file is generated automatically, inspect the output rather than assuming the generator selected the right pages. A long list of everything on the site can make a curated resource less useful.

When is creating one worth the work?

Create it when you can state who or what will use it, what information it will organize and how it will be maintained. Defer it when the site’s core service facts are wrong or the effort displaces a more consequential repair.

Google says its AI search features do not require special AI text files. Do not interpret publishing llms.txt as purchasing search eligibility. Our own citation sample points the same way: across 1,429 answers from ChatGPT, Perplexity and Google AI Overviews, the most-cited domains were reddit.com (657), baadigi.com (382), google.com (310), linkedin.com (291) and youtube.com (289). That sample measures cited pages, not files, and it rewards pages that answer a question plainly. The LLM SEO guide covers that page format; the citation study has the full table. The AI visibility check observes answers; that is a separate question from whether a resource file exists.

Our SEO service can help prioritize site work. If you need a wider review of one business and one market, BaaDigi’s AI Visibility Audit is $497. Its value should be a useful diagnosis, not an oversized invoice for a small text file.

Who should maintain the file after launch?

Assign it to the person who maintains the linked public information. When a service, policy or page changes, check its entry and destination at the same time. Remove stale links rather than preserving them to make the file look comprehensive. If nobody will maintain it, prioritize accurate ordinary pages first. Use the website task checklist to test the actual customer action instead of treating file publication as acceptance.