What is llms.txt, and should you bother? An honest 2026 verdict

The short answer for a local business in 2026 is no, not for Google, and probably not for the reason you have been sold. llms.txt will not rank you, will not get you cited, and will not undo a thin location estate. Here is what it actually is, the realistic upside, and who should bother.

What llms.txt actually is

llms.txt is a proposed convention, not a standard. It is a single plain-text file at the root of your site, a hand-written index in Markdown that points to the pages you most want a large language model to read. The pitch is that an assistant fetching your site can read one tidy file instead of crawling everything. It sits next to robots.txt in spirit, but robots.txt is honoured by the search engines that matter, and llms.txt is not honoured by anyone in particular yet.

What Google says, in its own words

Google has put this in writing. In its AI features and your website guidance, it states: “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.”

It goes further on llms.txt by name. Google says it is fine to keep such a file for other services or systems that use these files, and that doing so “won’t harm (nor help) your visibility or rankings in Google Search, as Google Search ignores them.” So the file is not a ranking factor, not a tie-breaker, and not a soft signal. For Google AI Overviews and AI Mode, which read your live pages the same way ordinary Search does, it is simply not in the loop.

Who reads it in 2026, and who does not

The honest position is that the big consumer AI assistants have not publicly committed to fetching it as a discovery file. As of 2026 the major AI vendors have not publicly committed to reading llms.txt as a way to discover or rank sites, and the assistants most people use answer local questions from search indexes and live page fetches, not from a curated index you wrote for them.

Where the convention does have traction is narrower and worth naming plainly. Developer tools, documentation platforms, and AI agents already working on your site are the audience that benefits, and the public recommendations that exist are for documentation and agent use rather than for local-business discovery. A coding assistant pointed at your docs can genuinely save tokens by reading an index first. A potential customer asking an assistant for a plumber in Lewes is not reaching your llms.txt at all.

The realistic upside

The upside is real but small, and it is not a ranking upside. If you maintain technical documentation, run a platform with an API, or expect agents to operate inside your site, an llms.txt file is a reasonable courtesy that can make those tools faster and more accurate. For a service business whose pages are about towns and trades, the upside is closer to zero. The work that moves you in AISearch is the same work that moves you in ordinary search: pages a person would find useful, real local substance, and clean structured data.

Who should actually bother

Bother if you fall into one of these cases.

• You publish documentation or a developer reference and want agents and tools to read an index rather than crawl the lot.
• You can generate and update the file automatically from your published content, so it never drifts out of date.
• You understand it is for those tooling cases and are not expecting it to do anything for Google.

Do not bother if any of these are true.

• You are a local or service-area business hoping it will help you rank or get cited. It will not.
• You would be writing and maintaining it by hand, where it will go stale the moment your pages change.
• You are reaching for it instead of fixing thin pages, missing local substance, or absent structured data, which are the things that actually move you.

If you do decide to keep one, treat it as the last item on the list. Generate it from your live content so it stays honest, and spend the saved effort on the page itself, as long as you understand it is doing nothing for Google right now.

How to build one, if you have decided to

Keep it boring and keep it generated. Place a plain-text file named llms.txt at the root of your domain. Open with a single heading for your site and one line on what you do, then list the handful of pages you would actually want an assistant to read, each as a Markdown link with a short description. Leave out the marketing, leave out anything that duplicates a page, and resist the urge to list everything. The whole value of the file is that it is short.

The part that decides whether it is worth having is maintenance. A file written once by hand is worse than no file, because it tells tools something that quietly stops being true. Build it from your published content so it regenerates when your pages do. That is the only version of llms.txt worth keeping.

The verdict

For a local business, skip it and lose no sleep. Google has told you in plain terms that it ignores the file, so there is no ranking case to make. Put the time into pages a person would value and into LocalBusiness structured data, which is the markup that genuinely helps machines understand a service-area business. If you want the wider picture of how assistants find and quote local firms, the guide on how AI search finds local businesses covers it. And if maintenance is your worry, an engine that generates and updates your local pages from real inputs, with the structured data handled, is the version of this work that actually pays off. The documentation sets out how that fits together.