What a scan sees.
When Limen scans a site, it reads the page the way an AI assistant does — no browser, no goodwill — and writes down what it could and couldn’t take. Below is the document that comes back: a complete, real report, rendered by the same code a live scan uses. Read it here at your own pace; nothing on this page is a screenshot.
Meet Atelier Flora.
A fictional florist studio in Zurich — bouquets, weddings, workshops, a lovely website that nobody ever tuned for machines. We invented the shop, then let the real engine judge the evidence: only the site is fiction; the arithmetic is exactly what your own scan runs. Her story continues through the coming chapters — this is the day the machines read her site for the first time.
The real report component, rendering the fictional scan of atelier-flora.example.
How the machines see atelier-flora.example
We read your homepage the way an AI assistant would — its text, its markup, its front door — and asked AI the three questions every customer asks. Here’s where you stand relative to the threshold of perception.
4 to fix. 7 already passing.
The fix. Put your services on the page itself as plain text, not as a document to download.
The fix. Update the © year in your footer and add one visible “last updated” line — two small edits that tell AI you’re still in business. Your full report shows exactly where they go.
The fix. Add a small block of structured data that states your name and what you are, in the format machines expect. Your full report includes the exact snippet, ready to paste.
Heading structure · Your services are hidden from AI
Your site looks abandoned to AI
Location in page text · Contact details · Page title & description · Meaningful page title · AI can’t confirm your basic details
AI crawler access · Content without JavaScript · llms.txt guidance file
Passing checks in grey, findings in red. + 8 deeper checks also ran — part of the paid product.
This is one model reading one page. The paid product tracks the prompts your buyers actually type, across models, scan after scan.
Top issues: Description length · Web app manifest · og:image dimensions
Full Open Graph report →How the score is built. Every point is a named check — weighted high 3, medium 2, low 1 — and the score is the weighted share that pass, per territory and overall. AI reads your page, but never scores it. The same scan run twice gives the same number.
Four things you just saw.
The report is written to be read without training. Still, four habits of the document are worth naming — they hold for every report Limen produces, including the one it would write about your site.
The score, and the threshold. The number at the top is the weighted share of passing checks. 80 is the limen — the threshold a site crosses when assistants can read it, trust it, and file it correctly. The bar shows exactly how far the shop stands from it.
Quick wins carry their fix. The report never stops at naming a problem: the top findings arrive with what to change, in plain words. No jargon survives, and nothing is diagnosed without a way to act on it.
Named checks, and an honest count. The free report names its own set of checks and admits the deeper ones only as a count. What you can see is real; what you can’t is never pretended. That rule holds everywhere in Limen.
Four territories. Every check belongs to Content, Trust, Identity or Access, and the rail scores each one — so you can tell a writing problem from a locked door, and spend your effort where the score actually moves.
Read your own site this way.
The same engine, the same honesty, on your address — about a minute, no account needed to look.
Free to look. No card, no call.