How it works

The process that puts your name in the answer.

01SCAN
15 pages · read like a machine
02READ
four territories · every check scored
03FIX
plain language · Janus prompts
04RE-SCAN
same checks · proof of change
Free to look · no card, no call
The process

Four moves from unread to named.

One loop does the whole job: read what the machines can see today, score the gap, hand you the fixes, and prove the change. No agency, no call — the software runs the moment you do.

01 · Scan

Read the site the way machines do.

Limen fetches up to 15 pages per scan — the homepage plus the pages you choose — and reads each one the way an AI assistant would: no browser, no JavaScript excuses, no brand goodwill.

02 · Read

Score what the machines see.

Every page is judged against the same fixed set of checks, grouped into four territories — Content, Trust, Identity, Access. One verdict per check, one score per page, one picture of the site.

03 · Fix

Fix with Janus at your side.

Every finding arrives twice: in plain language that says exactly what to change, and as a paste-ready AI prompt from Janus, Limen’s resident guide — drop it into your AI tool and the fix comes back written.

04 · Re-scan

Prove the change.

Re-scan whenever you’ve shipped — same checks, new verdicts. Watch findings flip from open to clear, check by check, and let the weekly digest keep watch between scans.

How the scan works

15 pages. Read the hard way.

The scan tells you what the machines can actually take from your site — not what your browser shows you. Limen reads your pages the way ChatGPT, Claude, Gemini, Copilot and Perplexity read them when they decide whether you’re worth citing.

You choose the pages that decide. The homepage plus up to fourteen more — services, team, prices, the pages a buyer’s question actually lands on.

It reads like a machine, not a visitor. Are AI crawlers allowed in? Does the page exist without JavaScript? Is there llms.txt guidance? The scan checks the door before it judges the room.

Every scan is current. Sites drift and models change their reading habits. Each scan re-reads everything from zero, so last month’s pass doesn’t hide this month’s problem.

ChatGPTClaudeGeminiCopilotPerplexity
15
pages read per scan
4
territories, every check scored
How the reading works

Four territories. Four honest questions.

Every check on every page belongs to one of four territories, and each territory answers one question about your site. A score per territory, a verdict per check — and every failed check points at its own fix.

Content72

Can AI read, extract, and want to cite what you say?

Openings that state the point, sections that answer what they promise, questions and answers the machines can lift whole.

Trust58

Does AI have reason to treat you as credible?

Named people, real credentials, third-party traces — the evidence a machine looks for before it repeats your claim.

Identity81

Can AI tell what & where you are?

What the business does, for whom, in which place — stated so plainly a machine cannot file you under the wrong thing.

Access64

Can AI reach and read the site at all?

Crawler permissions, rendering without JavaScript, llms.txt — the door the machines must get through before anything else counts.

The scores above are illustrative. Your board shows the real ones, territory by territory.

Inside the fix

Janus writes the prompt. You paste it.

Janus is Limen’s resident guide. For every finding he does two things: explains in plain language what to change and why the machines care — and hands you a ready-to-use AI prompt, written against your actual page, that you paste into your AI tool to produce the fix.

No jargon to decode, no brief to write. The prompt already knows your page, your finding and the shape of a passing answer.

Your opening never says what you doContent · high impact

The first paragraph of /services talks about passion and journeys. A machine reading it still doesn’t know you do physiotherapy in Zürich. Say it in the first sentence.

JANUS PROMPTCopy

Rewrite the opening paragraph of the page below so the first sentence states, plainly: what the business does, for whom, and where. Keep the warmth, cut the metaphor. Current opening: “Every body tells a story…”

Paste into ChatGPT, Claude or your site editor’s AI — the fix comes back written.
What you get to work with

Every finding, ready to act on.

The dashboard isn’t a diagnosis you file away. Each finding carries its own fix, sorted so the change that moves your score fastest sits on top.

Plain-language findings

What’s wrong, why a machine cares, and what to do — written for owners, not engineers. No acronym survives unexplained.

diagnosis · why · fix

Janus AI prompts

A paste-ready prompt per finding, built against your page. Your AI tool does the writing; you approve and publish.

copy · paste · publish

Site-wide patterns

When the same problem repeats across pages, Limen folds it into one pattern — one decision fixes eleven pages, not one.

one fix · many pages

The report & the brief

A PDF report for the drawer and a developer brief for the handoff — every finding, every fix, in the order to work them.

PDF · developer brief

The measurement loop

runs on your clock
Scan day
15 pages read, findings on the board
The week after
fixes shipped, Janus prompts pasted
Re-scan
same checks, new verdicts
Site score
5478+24
Checks passing
3143+12
Findings open
145−9
Territories clear
13+2

Illustrative deltas from a first fix cycle. Your board shows the real ones, check by check.

How the re-scan works

Same checks. New answer.

Progress you can’t measure is progress you can’t trust. The re-scan exists so the change is visible in the verdicts, not just in your effort.

An identical yardstick. The same checks, the same territories, the same scoring. The change is trustworthy because the test never moved.

Finding by finding, before and after. Each fix flips its own check from open to clear — you see exactly which move paid, on which page.

On your clock, not ours. Re-scan the hour you ship or the month after. Between scans, the weekly digest watches the site and flags what drifts.

Models change. The fixes hold.

Assistants are reweighted every quarter and new ones arrive without warning. Limen’s checks are model-agnostic — they test the structures and signals every major system learns from, not the quirks of one. When the machines change how they read, the checks are retuned so a site that passes today still passes tomorrow.

Model-agnostic checksRetuned as models moveFive assistants, one yardstick

Buyers ask machines. Be the answer.

One scan, four territories, a fix for every finding — and a re-scan that proves your name now stands where the silence was.

Free to look. Two plans to act.