How it works

1. We read the page the way a visitor would

We load the public page you give us, capture a screenshot, and pull out its visible text, headings, links and buttons. Hidden elements, closed pop-ups and template placeholders are removed so the panel only reacts to what a visitor would actually see. Behind a login? Upload a screenshot instead.

2. Thirty simulated visitors react

An AI model plays 30 distinct visitors, each with their own background, goal and level of patience, matched to the audience you describe (or a sensible default for the type of site). Each one says what they understood, what confused them, whether they'd take the next step, and why. Panelists are asked to be candid, not polite.

3. We summarize and rank fixes

The 30 reactions are combined into clarity and trust scores, the places the panel got stuck (with how many panelists mentioned each), and a ranked list of fixes. Any number a panelist quotes is checked against the page; figures that don't appear on the page are corrected or removed.

What is simulated

• The visitors. They are AI personas, not real people, and are not drawn from your analytics.
• Their reactions, scores and "would take action" answers.
• Photo votes: each simulated viewer picks a favorite or says none fit.

What isn't simulated

• Your page content and screenshot: the panel reviews the real page.
• Fact-checks on numbers and claims against that page.
• Before & After rebuilds: a real working page built from the top fixes.

How to use the results

Treat the panel as a fast, honest first read. It's good at spotting confusing copy, missing trust signals and unclear next steps. It doesn't replace talking to real customers, A/B tests or your analytics, and it can't measure real traffic or sales.