1 · See it
The agent view
The answer an agent receives about any domain, rendered for a person: each action’s decision and the reason.
llms.txt audit: does yours resolve, parse and match the site it describes?
In one sentence
An llms.txt audit checks whether the file at /llms.txt answers 200 to every crawler identity with a text content type, parses, and cites only pages the site actually declares and serves — a file that answers 200 with a login wall or an HTML shell counts as absent.
Type a domain. The scan fetches /llms.txt on the apex and www hosts as fifteen client identities, records the status, content type, cache age and bytes each one received, and checks the file against the pages the site actually declares.
Measured
What the scan measures for this
The file itself
Does /llms.txt answer 200 to every identity, with a text content type, from a cache that is not stale? A file that answers 200 with a login wall or an HTML shell is scored as unreadable, not present.
The surfaces around it
llms.txt, agents.md, the entity map and the sitemap are read together. A file that cites pages the sitemap does not declare, or that the site redirects, is a claim the site does not support.
Instruction surfaces
What the file tells an agent to do, whether those instructions came from you or from a template, and whether any of them contradict robots.txt.
See this section on a real report: Instructions for agents →
What an agent is served
The page an AI client receives, compared with the page a browser receives. llms.txt cannot compensate for a homepage that ships as a script shell.
For agents
What an agent is told about the same site
An agent does not read the report. It asks one endpoint before it reads, cites, connects or transacts, and gets a signed answer built from the same measurement.
2 · Call it
One endpoint
GET /api/v1/resolve?domain= returns the signed answer; preflight returns a decision and a receipt for one action.
3 · Check it
Without trusting us
Every answer is Ed25519-signed and logged; it verifies offline with the published key.
Workflow
The workflow
- Scan the domainThe free scan fetches the file and every surface around it. Nothing is installed and nothing changes on the site.
- Open the Sections tab and read Machine files and SurfacesEach row names what was requested, what came back, and the bytes that decide the row.
- Draft or correct the fileIf there is no file, the free draft below builds one from the pages the site already publishes. If there is one, the fix list names the lines that fail.
- Publish, purge the cache, re-scanDone is measured, not asserted. The second scan is the proof, and on a watched domain it becomes the first point of the record.
Free
Free tools for this
Each one is a single request to the domain you name, answered as text or a download. Put the domain on the end of the URL, or use the form on the tools page.
Draft an llms.txt from your own pagesHomepage title and description, then every declared page with its own. Cut it down: the value of the file is the choosing.Can a machine read your files?robots.txt, sitemap, llms.txt and the homepage, read by content type and cache age.
Reading
Read next
- How to write an llms.txt and verify it
- An llms.txt that answered 200 with a login wall
- The instructions on your domain you did not write
Start from what you are checking
robots.txt for AI crawlerswhat yours actually permitsEntity map validatordoes your entitymap.json parse, and does every claim in it resolve?SEO audit toolthe technical checks, plus what AI crawlers actually receiveTechnical SEO auditcrawl access, sitemap, hosts, links and speed, checked in one scanLocal SEO auditdoes your site say who you are, where, and does every listing agree?SEO audit reports for agenciesyour name on the report, the proof still attachedWhat is a website auditWhat is a website audit? What it checks, what it costs, and how to run one freeWhat is an SEO auditWhat is an SEO audit? The checks that matter, in the order to fix themAEO auditcan ChatGPT, Claude, Perplexity and Google reach, read and quote your site?
You cannot pay to improve a score. A licence adds the lists behind the counts and the record over time - never a better grade.
Questions about this page
QDoes an llms.txt improve AI visibility on its own?
No. It is one machine surface among several, and the scan weights the homepage a crawler actually receives far above it. A site with a perfect llms.txt and a script-shell homepage still fails the Read stage.
QWhat if my site has no llms.txt?
The scan records it as absent, not as a defect that caps the grade. The free draft tool builds one from your homepage and declared pages; the value of the file is choosing which pages to keep, so cut it down before publishing.
QWhy does the scan fetch it as fifteen identities?
Because a file served differently to GPTBot than to a browser is the defect that matters. Each identity's status, bytes and content type are recorded separately, and any divergence is a finding with the evidence bytes attached.
QIs this the same as the homepage scan?
Yes. It is the same free scan and the same report; this page opens it at the sections that answer the llms.txt question. Nothing here is a separate score.