llms.txt for B2B websites: what it is, what to put in it, and a generator

llms.txt is a proposed Markdown file at your site's root that gives AI agents a short guide to your most important pages. For B2B sites, include your name, a summary, core services, pricing and key proof pages. Be realistic: no strong evidence shows it earns citations, so it should take an hour at most.

What is llms.txt?

The llms.txt proposal came from Jeremy Howard of Answer.AI in September 2024, and the current page is marked v2, modified August 10, 2026. It proposes a small file that gives language model agents a curated guide to a site. The proposal contrasts it with robots.txt, which states access rules, and sitemap.xml, which lists too much to fit into context.

It is a proposal rather than a standard that search engines have adopted. The site describes its main use as on-demand help during inference, when an agent is helping a user. Training use is possible too, but the core idea is a reading list for an agent, not a ranking signal.

Do AI engines actually use it?

Here is the evidence, sorted by what it can and cannot tell you.

The same report says GPTBot occasionally fetches llms.txt files, without any tie to citation outcomes, and that OpenAI's crawler docs focus on robots.txt and do not say the file affects citations. So a fetch is not the same as a benefit.

What can be fairly concluded? On discovery and citations, nothing supports claims that llms.txt helps, and Google has explicitly said it is unnecessary for its AI features. Yet absence of evidence is not proof of zero effect, since SE Ranking's own wording was "at least not yet," and engines could change. The narrow case Mueller allows, an agent already on your site trying to complete a task, is the most plausible use today.

Should a B2B website bother?

Treat it as optional. A short file costs little, but it should never come before work with stronger support:

1. Crawl access for OAI-SearchBot and PerplexityBot.
2. Indexable pages with accurate, plain-text facts.
3. Answer pages for real buyer questions.
4. Placement on third-party sources.
5. Then, if you want, llms.txt.

Be wary of anyone selling llms.txt as an AI visibility fix, or charging separately for it. For transparency, Fame publishes its own file and offers a generator. Fame cannot show that the file contributed to any of its own AI citations, because the effect cannot be separated from everything else on the site.

What format does the file follow?

The spec sets out a simple Markdown layout, in this order:

• An H1 with the project or site name. This is the only required element.
• A blockquote with a short summary.
• Optional detail paragraphs or lists, with no headings.
• Zero or more H2 sections containing lists of links, each written as a Markdown link and optionally followed by a colon and notes.
• By convention, an H2 called Optional holds secondary links an agent can skip when short on context.

The file normally lives at yoursite.com/llms.txt, although the spec also allows a subpath such as /docs/llms.txt, which covers the URLs below it. When more than one applies, agents should use the most specific.

What should a B2B company put in it?

Keep it factual and small. Here is a line-by-line template. Acme is an invented company. Replace each example with verified facts about your own, and keep the Markdown symbols shown in the first column.

A few rules for B2B sites:

• Cap it at around 20 to 30 links. The purpose is curation, and this cap is a judgment call rather than part of the spec.
• Write notes that describe the page. Avoid promotional adjectives, since Mueller's main objection is that the file is self-reported.
• List public pages only. Leave out login, staging, and anything private. The file is public.
• Keep every claim consistent with what the linked page says. Include prices only if they are already published.
• Name who it reviews and updates it. Put a calendar reminder in for each quarter, or generate it from your CMS if that is simple.

Do you need the Markdown versions too?

The proposal also suggests clean Markdown copies of pages, at the same URL with .md added or swapped in, along with link relations that help clients find them. That is real engineering work on most marketing sites, and Mueller has called building Markdown pages for bots "a stupid idea." He also estimates agent-navigation standards could take six months to a year or more to settle.

For a typical B2B site, skip this and link to your normal pages, unless you have developer documentation where agents and developers will use it. Docs are the more natural fit, since an agent helping a developer on your site is close to the narrow on-site use Mueller allows.

How do you create it quickly?

Use this short process:

1. Open Fame's llms.txt generator and create a first draft, or write it by hand from the template above.
2. Delete everything except your 20 to 30 most useful links.
3. Check each link opens and that the description is accurate.
4. Upload the file to your site root so it loads at yoursite.com/llms.txt.
5. Confirm it loads as plain text in an incognito browser window.
6. Check your server logs after a few weeks if you want to know whether anything fetches it. Log hits show interest rather than benefit.

Then go back to the work that has stronger support. How to get cited by ChatGPT and how to write an answer page are better uses of your next hour.

If a broader program is what you are after, Fame's LLM SEO service is part of Fame Search, a paid service at $2,000 a month. An llms.txt file alone would be a poor reason to buy it, or any similar service.