Paste your site and get a draft llms.txt built from your home page and up to ten other pages, ready to copy or download, with every gap marked instead of guessed.
Reads up to eleven public pages. Takes up to ten seconds. Nothing is stored.
Many llms.txt generators ask an AI model to write the descriptions. That reads well and is often wrong: a model that does not know your business fills the gaps with plausible guesses, and those guesses then sit in the one file meant to be a reliable summary of your site.
This one copies. Each line comes from the page it links to: its title or main heading, its meta description or first paragraph. Your name comes from your markup or your og:site_name. Your phone and email come from the tel: and mailto: links on your home page. What is not there stays visibly empty.
# Harbor Dental Studio > Family and cosmetic dentistry in Portland, Oregon: check-ups, implants and same-day crowns. ## Services - [Dental implants](https://example.com/implants/): what an implant costs, how many visits it takes, who is a candidate - [Same-day crowns](https://example.com/crowns/): one-visit crowns made in the clinic ## Contact - Phone: +1 503 555 0100 - Address: 12 Harbor St, Portland, OR
An invented example. One heading with the name, one line saying what you do and where, then sections of links with a line each.
llms.txt tidies your site for machines; it does not make an assistant recommend you. The free OperStack check scores how readable your site is and asks ChatGPT your buyers' questions to see whom it names.
A plain text file at the root of a site, /llms.txt, proposed at llmstxt.org. It starts with one heading naming the site, a one-line summary, and then sections of links, each with a line on what the page is. It is a map written for programs rather than people.
Do not count on it. In our study of 731 websites, ChatGPT named sites with an llms.txt in 32% of cases and sites without one in 36%: no noticeable difference. The file keeps a site tidy for machines that choose to read it. It does not decide who gets named.
Not for Search. Google says you do not need new machine-readable files or AI text files to appear in AI Overviews and AI Mode (Google: AI features and your website). Rankings are not affected either way.
Only from your pages: the title or main heading of each page, its meta description or first paragraph, and contact details from your markup, phone and email links. Nothing is written by an AI model and nothing is guessed. Where a fact was not on the pages, the draft says [fill in: ...] so you can see what is missing.
Your home page and up to ten more, taken from your sitemap or, if there is none, from the links on your home page, with no more than three from one section so the draft shows the whole site. Twelve requests at most, all within about eight seconds. A big site needs a fuller file: the free OperStack Astro starter builds one from the whole site on every deploy.
No. The draft is built while the page loads and shown only to you. The download button makes the file in your browser. Nothing is written down on our side, and there is no account or cookie.
Built by Maksim Shchegolev, founder of OperStack. Free, no account, no cookie. The address you check is not stored.
OperStack