Can an AI agent read your site?
Nine checks on the public configuration of any domain: crawler rules, structured data, sitemap, a machine-readable copy. Free, no signup, a shareable link at the end.
Reads the homepage, robots.txt, sitemap.xml and llms.txt. Nothing else, and nothing is stored about you.
- Checking
robots.txt reachable
Without one, crawlers get no guidance at all.
- Checking
AI crawlers allowed
A blocked crawler reads nothing at all, which is why this carries the most weight.
- Checking
Sitemap declared
It is how a crawler learns which pages exist without guessing from links.
- Checking
Structured data on the homepage
Structured data is the part of a page a machine reads without interpreting prose.
- Checking
Organization schema
It is what makes the site a resolvable entity rather than an unattributed document.
- Checking
Machine-readable copy
A plain-text copy removes the parsing step between the content and the model.
- Checking
Title and meta description
Cheap to fix, and it is what a summary is assembled from when nothing better exists.
- Checking
Single h1
Zero leaves the page with no stated subject, and several leave it with several.
- Checking
Content signals
Optional and emerging. A missing one is a small deduction, not a problem.
What the AI-readiness check is
The AltoRank AI-readiness check is a free test of one domain’s public configuration. It fetches four
URLs on that domain, the homepage, /robots.txt, /sitemap.xml and /llms.txt,
and returns nine results, each one pass, fail or unknown, with the evidence it saw and the fix for a fail.
It never crawls past the homepage, it needs no account, and it does not need you to own the domain.
The nine are the technical preconditions for an AI assistant reading a site at all: permission in robots.txt, a map of the pages, structured data saying what the organisation is, and a plain-text copy of the content. They are a precondition, not a promise. A site can pass all nine and still be cited by nobody, because being readable and being worth quoting are different problems.
Each check carries a weight of 3, 2 or 1, totalling 18, and the score is the share of that weight earned by the checks that completed. A check that could not be decided is reported as unknown and left out of both sides of that fraction, so a slow server lowers confidence rather than the score.
For context on the number you get back: across 274 European agency sites scanned on 16 August 2026, the median score was 83 and 6 sites passed all nine. The per-check pass rates are in the AI Readability Report 2026.
The nine checks, in full
In the order they are run and reported. Nothing else is tested, and nothing here is inferred: each row is decided from what the site served at the moment of the check.
| Check | What it tests | Weight |
|---|---|---|
| 1. robots.txt reachable | Whether /robots.txt returns a readable file. Without one, crawlers get no guidance at all. | Medium (2) |
| 2. AI crawlers allowed | Whether robots.txt blocks any of eight named AI crawlers, tested one by one: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Google-Extended, CCBot, Applebot-Extended. A blocked crawler reads nothing at all, which is why this carries the most weight. | High (3) |
| 3. Sitemap declared | Whether a Sitemap: line in robots.txt points at a sitemap that returns 200, or /sitemap.xml does. It is how a crawler learns which pages exist without guessing from links. | Medium (2) |
| 4. Structured data on the homepage | Whether the homepage HTML contains any JSON-LD, however deeply nested. Structured data is the part of a page a machine reads without interpreting prose. | High (3) |
| 5. Organization schema | Whether that JSON-LD includes an Organization or LocalBusiness type. It is what makes the site a resolvable entity rather than an unattributed document. | High (3) |
| 6. Machine-readable copy | Whether /llms.txt exists, or a markdown version of the page is offered. A plain-text copy removes the parsing step between the content and the model. | Medium (2) |
| 7. Title and meta description | Whether the homepage has both a title and a meta description. Cheap to fix, and it is what a summary is assembled from when nothing better exists. | Low (1) |
| 8. Single h1 | Whether the HTML the server sends contains exactly one h1. Zero leaves the page with no stated subject, and several leave it with several. | Low (1) |
| 9. Content signals | Whether robots.txt carries a Content-Signal line stating ai-train, search and ai-input preferences. Optional and emerging. A missing one is a small deduction, not a problem. | Low (1) |
Total weight 18. A domain passing the six checks worth 12 points scores 67 out of 100.
What this means
A fail is something the site owner can change today: a line in robots.txt, a block of JSON-LD in the head, a text file at the root. An unknown means the server did not give a clear answer and the check claims nothing about it. The score weights crawler access and structured data above titles and headings, because a blocked crawler reads nothing at all.
Every result gets its own page at app.altorank.co/check/yourdomain.com, which
can be sent to whoever owns the site, and a one-line script that renders the current ratio and date as a
badge on a page of your own. Both re-read the same result, so neither can quietly go stale against the
other.
AltoRank runs the same nine checks on every site you add and generates the schema and llms.txt for you to review before anything is published. The checker is part of the open-source repository, so the scoring can be read rather than trusted.
Frequently asked questions
What does the check read?
Four URLs on the domain you give it: the homepage, /robots.txt, /sitemap.xml and /llms.txt. It reads public configuration only and never goes deeper than the homepage.
What does unknown mean?
The server did not give a clear answer, for example a 503 on robots.txt, or the site was slow and that check did not run before the 25 second deadline. Unknown is not a fail and does not lower the score.
How is the score calculated?
Each check carries a weight of 3, 2 or 1. Weight 3: AI crawlers allowed, structured data on the homepage, Organization schema. Weight 2: robots.txt reachable, sitemap declared, machine-readable copy. Weight 1: title and meta description, single h1, content signals. That is 18 points in total, and the score is the share of that weight earned by the checks that completed, rounded to a whole number. A site passing six checks worth 12 points scores 67.
Is the check free?
Yes. There is no account, no email gate, no daily quota and no paid tier of it. An email address is optional and is used once, to send you the result you are already looking at. There is no list.
Do I need to connect anything?
No. The check reads only what your server already serves to the public, so there is no tag to install, no analytics or Search Console property to connect, no DNS record and no access to grant. You do not need to own the domain to check it.
Do I need an account?
No. Type a domain and the result appears here, with a link you can share. An email address is optional and only used to send you that result. Adding the domain to AltoRank is a separate step, and self-hosting is free.
Can I run it on a client or prospect domain?
Yes, on any domain, and without limit or attribution. The result has its own shareable page, so it can be sent to the person who owns the site.
Does a good score mean AI assistants will cite me?
No, and nothing here claims that. These nine checks measure whether a site can be read, which is a precondition rather than a cause. A perfectly readable site with nothing worth quoting still gets quoted by nobody.
Fix what it finds
Every fail on the list above is something a site owner can change today, usually in one file. These are the free tools and written methods for doing it by hand, all of them on this site and none of them behind a signup.
AI Readability Report 2026
The same nine checks run across 274 European agency sites, with the pass rate for each one. What a score means, in context.
robots.txt generator
Build a robots.txt with explicit AI crawler rules, for the first and second checks.
JSON-LD schema generator
Generate valid Article, FAQPage and BreadcrumbList markup, for the structured-data checks.
SEO health check
The longer manual audit this automates one slice of, ordered by what actually blocks ranking.
Internal link checker
Paste a page and see every internal link and anchor, which the nine checks do not cover.
Free growth plan
The other hosted check: what a domain ranks for, what is within reach, and what is missing.
The rest are indexed on the free SEO tools hub. If you are picking a tool rather than fixing a file, the tool comparisons are written for that, and agencies running this on a client roster will want the agency page.
Be the site the assistant names.
Add a domain and the first draft is written while you watch. Every publish is your decision, a click on the draft or a rule you set and can hold, and you can read the source or self-host it free.
Add a domain, it sets up your workspace