DefinitionA proposed plain-text file at the root of a website that gives large language models a curated overview of the site's most important content, similar in spirit to robots.txt for crawlers.
llms.txt is a proposed plain-text file at the root of a website (yourshop.nl/llms.txt) that gives large language models a curated overview of the site’s most important content, similar in spirit to robots.txt for search engine crawlers. The proposal came from Jeremy Howard of Answer.AI in September 2024. The file is written in markdown: an H1 with the site name, a short summary in a blockquote, and sections of links to key pages, each with a one-line description. An optional companion, llms-full.txt, holds the full text of those pages in one document.
The misunderstanding I see most often is that the file makes your shop show up in ChatGPT, Gemini or Claude. Nobody has shown that it does. This is a proposal, not a standard. As of 2026, none of the major AI providers (OpenAI, Anthropic, Google) has officially confirmed using it for ranking or retrieval. Google’s John Mueller compared it to the old keywords meta tag: a file in which site owners describe themselves, which is exactly why systems tend to ignore it. Treat it as a low-cost experiment, not as an AI visibility strategy.
Why llms.txt matters (less than you think)
The idea behind the proposal is sound. HTML pages are full of navigation, cookie banners, scripts and product filters, and a language model has to wade through all of that to find the content. A clean markdown summary that says “this is a Dutch webshop for home batteries, here are the ten pages on range, pricing, installation and warranty” is easier to use. If an AI agent ever fetches your site live and looks for the file, you have given it a good starting point for free.
For a webshop the honest consequence is this: there is no revenue, click or conversion effect that I can attribute to an llms.txt file today. What moves AI visibility is being cited, and citations come from the same things that drive organic rankings: clear product content, strong category pages, mentions on other sites, and direct answers to the questions people actually ask. That is the territory of generative engine optimisation and answer engine optimisation. The file is a footnote to that work, not a substitute for it.
How to check whether llms.txt does anything for your site
First, check that the file exists: open yourshop.nl/llms.txt in a browser and confirm it returns a 200 status with plain text, not a themed 404 page or a redirect to the homepage. I run the same check in Screaming Frog with a list crawl, which also tells me whether the links inside the file resolve. Links to deleted pages are worse than no file at all.
Second, look at your server logs. That is the only place you will see whether AI crawlers (GPTBot, ClaudeBot, PerplexityBot) actually request it. In the logs I have reviewed for clients, hits on the file are rare and mostly come from SEO tools and curious competitors. For the effect you care about, citations and referral traffic, I use GA4 to segment sessions by referrer (chatgpt.com, perplexity.ai) and Ahrefs Brand Radar to track where a brand gets mentioned in AI answers. Neither changes because you added a text file, which is itself the finding.
Common llms.txt mistakes
Treating it as the AI strategy. Shop owners read a LinkedIn post, add the file, and expect to appear in ChatGPT answers the following week. Nothing happens, and they conclude that AI visibility is a myth. The file was never the lever.
Dumping the whole catalogue in it. An llms txt file with 4,000 product URLs is not curated, it is a second sitemap. The proposal is explicitly about the small set of pages that explain who you are and what you sell: category pages, buying guides, shipping and returns, your about page. You already have an XML sitemap for completeness.
Letting it go stale. A file generated by a plugin that was installed once and never updated keeps listing pages you deleted and prices you changed. If you cannot commit to checking it twice a year, keep it short so there is less to go stale.
Blocking the bots you are trying to inform. I regularly find shops that added the file while their robots.txt disallows GPTBot and ClaudeBot, or a hosting firewall blocks AI user agents. Decide first whether you want AI crawlers on the site at all.
How I handle llms.txt for clients
My rule is simple: if it takes 30 minutes, add it; if it takes longer, something else deserves the time. On WordPress and WooCommerce there are plugins that generate the file from your pages and categories, and several SEO plugins now include it as an option. On PrestaShop, Shopify and Lightspeed I write a static file by hand, 15 to 25 links with a one-line description each, served from the root. I write the descriptions myself rather than copying meta descriptions, because a model benefits from plain, specific language (“installation guides for home batteries in the Netherlands, including subsidy rules”) more than from marketing copy.
The deliverable is the file itself, a line in the technical SEO audit confirming it exists and resolves, and a sentence in the client report that sets expectations: a cheap bet with no proven payoff. The real AI visibility work sits elsewhere in the plan: content that answers questions directly, structured data, and mentions on sites the models already trust.
Frequently asked questions
Does Google use llms.txt?
No, not as far as Google has stated. John Mueller has said that none of Google’s AI services use it, and he compared it to the keywords meta tag. It has no effect on your normal search rankings and no confirmed effect on AI Overviews or AI Mode.
What is the difference with robots.txt?
robots.txt tells crawlers what they may not fetch, and every major crawler respects it. The other file tells language models what is worth reading, and no major provider has committed to using it. One is an established access control, the other is a suggestion. You need the first; the second is optional.
Should I also create an llms-full.txt?
Only if you can generate it automatically. It contains the full text of your key pages in one markdown document, which suits documentation sites and is rarely worth maintaining by hand for a webshop. A short, accurate file beats a long one that drifts out of date.
Can the file hurt my site?
Not directly. The only real risks are links to broken pages, descriptions that contradict the site, or the false sense that your AI visibility is taken care of. Keep it short and accurate, keep your expectations low, and the downside is close to zero.