LLMs.txt Checker
Get mentioned by AI assistants
MentionAgent helps your brand get cited by ChatGPT, Claude, and other AI models through strategic link building and content placement.
Start Getting Mentioned On AutopilotWhat this tool validates
| Check | Severity | What it means |
|---|---|---|
| H1 heading present | Required | Your file must start with a # Site Name heading to identify your site. |
| Blockquote description | Required | A > description line gives AI models a quick summary of your site. |
| Contains links | Warning | At least one [text](url) link helps AI models find your key pages. |
| Section headings | Warning | ## Section headings organize your content into logical groups. |
| Valid URL format | Warning | All links should use valid URL patterns starting with http:// or https://. |
| No HTML tags | Warning | llms.txt should be plain Markdown, not HTML. |
| File not empty | Required | An empty file provides no value to AI models. |
| Reasonable length | Warning | File should be long enough to be useful but not excessively large. |
Do the sites AI assistants cite actually have an llms.txt?
Every guide to this file says the same thing: add one, models like it. We checked whether that matches what cited sites do. On September 14, 2026 we took the 500 domains that ChatGPT, Claude and Gemini cited most often across the 8,635 answers in our citation corpus (collected July 8 to September 14, 2026) and requested /llms.txt on each, following redirects, with the same checks this tool runs.
What /llms.txt returned | Domains | Share |
|---|---|---|
| A valid file (plain text, opens with an H1) | 196 | 39.2% |
| Nothing (404 or another error status) | 275 | 55.0% |
| An HTML page (catch-all route or soft 404) | 20 | 4.0% |
| Plain text with no H1 | 6 | 1.2% |
| Could not connect within 8 seconds | 3 | 0.6% |
Three details matter more than the headline number.
- The biggest cited sites do not bother. The five most-cited domains in the corpus, reddit.com, medium.com, youtube.com, facebook.com and apps.apple.com, have no valid file; reddit answers the path with an HTML page. Among the top 100 cited domains, 43 have one. Among ranks 101 to 500, 153 of 400 do. Adoption is a mid-size-software-company habit, not a platform habit.
- Sites cited by all three engines adopt it more. 89 of the 188 domains cited by ChatGPT, Claude and Gemini serve a valid file (47%), against 34 of the 118 domains cited by only one engine (29%). That is correlation: larger, more technical sites both get cited more and adopt new conventions faster. Nothing here shows the file causing a citation.
- Many "valid" files are thin. Of the 196 valid files, 27 contain no links at all, just a heading and a description, and 31 skip the blockquote description the proposal calls required. The median file carries 37 links and about 10 KB of text; 42 domains also serve an
llms-full.txt.
The domains that pass include zapier.com, sproutsocial.com, buffer.com, hootsuite.com, semrush.com, blog.hubspot.com, g2.com, adobe.com, monday.com, mailchimp.com, notion.com, github.com and typeform.com. The corpus skews toward our customers' niches (social scheduling, sobriety apps, real estate education, scraping tools), so read the list as a sample of what gets cited in those markets, not of the web.
llms.txt format reference
The llms.txt file uses a simple Markdown-based format. Here is the basic structure:
# Your Site Name > A brief description of what your site does. - [Homepage](https://example.com) ## Key Pages - [About](https://example.com/about): Learn more about us - [Pricing](https://example.com/pricing): Plans and pricing - [Blog](https://example.com/blog): Latest articles ## Topics - Topic One - Topic Two
The H1 heading and blockquote description are required. Section headings (##) and links are recommended. The file should be plain Markdown with no HTML tags. Place the file at yourdomain.com/llms.txt.
Frequently asked questions
What is llms.txt validation?
llms.txt validation is the process of checking whether your llms.txt file follows the correct format and includes all required elements. A validator checks for proper Markdown structure, required sections like the H1 heading and description blockquote, valid link formatting, and overall content quality.
What makes a valid llms.txt file?
A valid llms.txt file must include an H1 heading (# Site Name) and a blockquote description (> description). It should also contain links in Markdown format [text](url), section headings (## Section), and be written in plain Markdown without HTML tags.
Is llms.txt required for AI visibility?
No. llms.txt is a community proposal, not a standard any assistant has committed to, and Google has said its search systems do not use it. What we can measure is adoption among sites that do get cited: 196 of the 500 domains cited most often across 8,635 ChatGPT, Claude and Gemini answers in our corpus serve a valid llms.txt, 39%. The five most-cited domains (reddit, medium, youtube, facebook, apps.apple.com) do not. Treat it as a cheap, harmless file to get right, not as a ranking lever.
How often should I update my llms.txt?
Update your llms.txt whenever you make significant changes to your site, such as adding new key pages, changing your site description, or restructuring your content. A good rule of thumb is to review it quarterly or whenever you update your sitemap.
What is llms-full.txt?
llms-full.txt is the optional companion file from the same proposal: instead of an index of links, it carries the full text of the pages you want a model to read, concatenated into one Markdown document. It is far less common. Of the 196 AI-cited domains in our sample that serve an llms.txt, 42 also serve an llms-full.txt. This checker validates llms.txt only.
Why does the checker say my domain returns HTML?
Your server is answering /llms.txt with a normal web page, usually a catch-all route or a soft 404, instead of a plain-text Markdown file. 20 of the 500 most-cited domains do exactly this, including reddit.com, forbes.com and linkedin.com. A model fetching the file gets your homepage markup, which is worse than a clean 404. Serve the file as text/plain or text/markdown, or remove the route.