Your website, as AI actually reads it.
Your buyers ask ChatGPT before they ask you. Give us your URL and we show you your homepage the way a machine sees it, every check it fails, and the files to fix it, ready to paste.
Free. 17 checks. The full report opens here and lands in your inbox.
We will email you this report and the occasional note about findability. Unsubscribe in one click, any time.
You can be invisible to AI with nothing on your site looking broken.
Your homepage loads fine. It looks finished. That tells you nothing about what a machine gets when it comes to read you, and the machine is what your buyers are asking.
Two things go wrong most often. Your pages build themselves in the browser, so a crawler receives an empty shell instead of your pitch. Or a security setting somebody switched on months ago quietly turns the crawlers away at the door, while your robots.txt still reads perfectly clean. Other tools count how often AI mentions you and cannot tell you why the number is zero.
What it checks
Four things, in the order they have to work. Every finding comes with the fact behind it, so you can check any of them yourself in a browser tab.
01Reach
+
If they cannot get in, nothing else on this page matters. Sites get locked out by a security setting somebody switched on months ago and forgot about. Your robots.txt can look perfect while it is happening, which is why we knock on the door as a bot and see what comes back.
- ·robots.txt rules for OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, GPTBot, Google-Extended and five more, separating the ones that gate live answers from the ones that only affect training data
- ·Live fetches of your homepage wearing the GPTBot and OAI-SearchBot user agents, compared against the same fetch as a browser
- ·noindex directives in meta tags and x-robots-tag headers
- ·Sitemap reachable and parseable
02Read
+
Most AI crawlers do not run JavaScript. If your page builds itself in the browser, they get the empty shell and leave. Your site looks finished to you and blank to them. This is the most common failure we see, and the one owners suspect last.
- ·Word count of the text in the served HTML, with no JavaScript run, reported next to the text-to-markup ratio
- ·Title tag present and descriptive
- ·Exactly one H1
- ·Meta description present and long enough to be useful
03Understand
+
A machine can read every word on your homepage and still not know what you sell. Clever headlines do not survive the trip. One plain sentence saying what this is and who it is for does more here than a rewrite of everything else.
- ·A plain sentence of 8 to 40 words stating what this is and who it is for
- ·JSON-LD present, valid, and typed
- ·Organization and Product or Service schema
- ·Headings phrased as questions, answered directly underneath
- ·llms.txt, reported but deliberately not scored: no AI company has committed to reading it yet
04Cite
+
AI repeats sentences that carry a fact. Pages full of adjectives get read and skipped. "Fast" is not quotable. "Ships in 72 hours" is. This gate asks whether anything you wrote is worth repeating.
- ·Machine-readable dates (schema dateModified, time elements, article:modified_time)
- ·Canonical URL
- ·Density of specific quantified claims in the page text
- ·Open Graph tags
The check returns one of five verdicts
Questions
Is the AI visibility check free?+
Yes, and it stays free. The scan makes no AI model calls and uses no paid services, so it costs us nothing to run however many people use it. We ask for an email because the report lands in your inbox as well as opening on the page, and because we would rather know who we built it for.
What do I get from the check?+
A score, the first thing to fix, the raw text a crawler receives from your homepage, and a plain-language instruction for every problem found. Then the files for whichever of those you are missing: structured data, a starter llms.txt, the exact robots.txt lines, written out from what is already on your page and ready to copy. If you already have all three you get none of them, which is the right outcome. One button turns the whole thing into a work order for whoever builds your site, and we email you a private link to the report so you do not have to keep the tab open.
How is an AI visibility check different from an SEO audit?+
SEO tools grade you against ranking best practices. AI visibility trackers count how often you get mentioned. This sits underneath both, because a mention count of zero tells you nothing about the cause. If a crawler only ever received 40 words from your homepage, that is the cause.
Which AI crawlers does the check test?+
It reads your robots.txt rules for OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Perplexity-User, GPTBot, Google-Extended, CCBot, Applebot-Extended, Bytespider and meta-externalagent. It then makes live requests wearing the GPTBot and OAI-SearchBot user agents, which catches firewall blocks that robots.txt cannot show you.
Does blocking GPTBot make me invisible in ChatGPT?+
No, and this is the thing most tools get wrong. GPTBot collects training data. OAI-SearchBot is the one OpenAI says decides whether you appear in ChatGPT search answers, and Google-Extended only affects Gemini training rather than Google Search or AI Overviews. The report tells you which of the two kinds of block you actually have, because they cost completely different things.
Does it check my whole site or just one page?+
The score and the four gates judge the one URL you give us, so the finding is always about a page you can go and look at. On top of that we follow up to eight more pages, taken from the links on that page first and topped up from your sitemap, and count the words a crawler receives from each. That is how you find out whether the problem is one page or the way the whole site is built. The robots.txt and firewall checks are site-wide by nature.
Why does my site score low when it looks fine in a browser?+
Most AI crawlers do not run JavaScript. A page that assembles itself in the browser can look finished to you and hand a crawler almost nothing. The report shows the exact word count a crawler receives, and the text itself, so you can see the gap rather than take our word for it.
Can I check a website I do not own?+
Yes. Everything here reads files that are already public: your HTML, robots.txt, sitemap.xml and llms.txt. It is what any crawler fetches anyway, so competitors work fine.
What if my score comes back high?+
Then stop working on this and go look somewhere else. Findability is not your problem. Traffic that shows up and does not buy is a positioning problem or a product problem, and both cost you more than this one.
Are the results stored?+
A scan is cached for a few hours so running the same URL again is instant, then it expires. We keep the result alongside the email that requested it. We do not publish scans and there is no public list of scanned sites. If you press "get a shareable link" you create a public page for that one result yourself, and only then: it shows the verdict, the crawler view and what was flagged, never your fixes or generated files.
Will you email me again after the report?+
Once. Four days later we send one follow-up: we re-run the check, and we name the single finding we would actually put in front of a developer first. That is the whole sequence unless you reply and ask for more. Every email has one-click unsubscribe.
Why does the check not just ask ChatGPT about my brand?+
Because that costs money every time someone runs it, and a tool with a bill attached has to either charge you or throttle you. We would rather keep this one free. Reply to the report email and we will ask by hand and send you the screenshot.
Findability is one of three layers we fix. The other two are positioning and product. How we work on findability.
A score is a symptom. The audit names the cause.
This check reads four gates on one page. The Broken-Layer Audit reads the whole business and says which of the three layers is costing you growth: positioning, product, or findability. 72 hours, written verdict, fix plan. Credited in full toward any sprint.
Nothing materially broken, no charge. Full audit details → · Sprint pricing →