AI website tool
Language models answer questions about what you offer before anyone opens your site. An llms.txt decides what they say. Enter your address, get the finished file, upload it.
How it works
No account, no installation, no follow-up. You get a text file that you check yourself and upload yourself.
The home page is enough. The creator finds the subpages itself, first from your sitemap, otherwise from the navigation.
Up to eight pages, each with title, description, headings and body text. Legal pages and the basket stay out.
Claude turns that into an llms.txt following the llmstxt.org proposal. Every link is checked against your domain.
Copy or download the file and place it as llms.txt in your root directory. Reachable at your-domain.com/llms.txt.
The difference
Without the file the model guesses from navigation and chance hits. With the file it reads your own framing first.
The model collects whatever sits at the top: menu items, the cookie banner, the footer. What actually makes your offer is often on page three.
“A provider in the services sector. No further information available.”
The model reads one sentence of context, your most important pages with a description each, and the terms that carry a special meaning at your company.
“Finished AI website templates to download, with a design system and seven AI skills, for freelancers, agencies and companies.”
Questions
A markdown file in the root directory of your website that tells a language model in one go what your site is about and where the important content sits. The proposal comes from llmstxt.org and works like a robots.txt, but for understanding rather than access.
Language models answer more and more questions before anyone opens a website at all. Being summarised wrongly there costs you customers without you noticing. An llms.txt sets out which pages count and how your offer is described.
Every request reads up to eight pages and puts a language model to work, which costs us real money. One request per day keeps the tool free and open to everyone. If you look after several websites, simply come back on another day.
No. The creator works only with the text it finds on your pages. Every URL in the finished file is checked against the list of pages actually found, and anything else is removed before you see the file.
It is used for the request and not stored. For the daily limit we keep only two short hashes with a salt that changes every day, from which neither your IP nor your device can be recovered. By the next day they are worthless.
The creator reads up to eight pages. For a shop with a thousand products it delivers the structure, not the catalogue. The file is then a good start that you extend with your most important categories.
The file is one of six signals that show whether a language model can do anything with your website. Whether the other five hold up, and whether your copy carries an answer at all, is what the AEO check shows.
Each answers one question about your website, in under a minute, without an account.
Website, the guideline and seven skills sit ready in one ZIP file. Pick your Template and start today, for yourself or your clients.
Or we build it for you: see the AI Website Service
Reference
The file follows a proposal from llmstxt.org. It is Markdown, it sits in the root directory, and it has a fixed order. Once you know it, you can tell in ten seconds whether a file is any good.
The first line starts with a hash and carries the name of the website, the way you want to be called. No slogan, no legal form, no city. A model uses this line as an identifier, and an identifier with seven words never appears in an answer.
Directly below sits a line with a greater-than sign. That is how you place yourself in one sentence, and it is the line most likely to appear word for word in an answer. It answers what you offer and who for. Two clauses are enough, three are too many.
Between the blockquote and the first heading you may put a paragraph. It is the only place for context, so for what a person would ask first in an introductory call. Directories and bullet lists do not belong here, the sections take care of those.
Every section starts with two hashes. Below it, each line holds a link in square brackets, then a colon and a half sentence saying what is on that page. Without that half sentence the link is worthless, because a model decides from it whether to open the page at all.
A section headed Optional carries a special meaning. It tells a model it may skip these pages when space runs short. Legal notice, privacy policy and archive pages belong there, your offer does not.
The standard is called llms.txt, with an s before the dot. Anyone who uploads llm.txt has a file nobody requests. After uploading, open the address in a browser and check that it shows the text and not your 404 page.
The three files do different things and do not exclude each other. robots.txt allows or forbids the request. sitemap.xml lists every address for complete coverage. llms.txt places and weights, and it is the only one of the three that has an opinion.
llms.txt is a proposal, not a standard with a committee behind it. Anthropic, OpenAI and Perplexity read parts of it, Google does not use it as a ranking signal. The benefit is that a model summarising your site finds a template instead of guessing from your navigation and footer.
An llms.txt with last year’s prices is worse than none, because a model states the old number with full confidence. Set a reminder for the same day you change your pricing page, and read the file through from the top.
34 finished industry websites, each with built-in knowledge and seven skills. Structure, copy, structured data and llms.txt are set before you change a single thing.
One payment, everything included. We show this at most once.
We would like to understand how this site is used with Google Analytics, loaded via Google Tag Manager. Nothing loads and no cookie is set before your ok. You can change your choice anytime in the privacy policy.
Choose what may be active. You can change your selection anytime in the privacy policy.