LandercheckThe comprehension score for landing pages

How the score works

The comprehension score answers one question: would a stranger understand what your page sells, in ten seconds and after a scroll? Here is exactly how we get there, and what it can't tell you.

What we measure

Not whether your page is good-looking, follows best practice or ranks in search. Only whether strangers can say what it is, who it's for, what they'd get, what it costs and what the button does, and how quickly they can say it. It's modelled on the five-second test that researchers run with real people.

The five visitors

Each visitor is an AI with a different goal and reading depth, on a different model, so no single reader's habits decide your score. For business pages, the feed scroller becomes a busy decision-maker.

Feed scrollerClaude HaikuBig words and pictures only; leaves fast
Total newcomerClaude SonnetLiteral; finds jargon and missing basics
ScepticClaude OpusReads everything; hunts the real price and the catch
Comparison shopperClaude SonnetAlready pays for something similar; wants the difference
BuyerClaude OpusBecomes the person the page targets; ready to buy if it's clear

The three stages

We test on a phone first, because that's where most paid and social traffic lands: a tap from a feed, on a small screen, with a thumb ready to go back. We load your page at iPhone size and check the desktop view at the end. Each visitor only gets the next stage after answering the current one, and earlier answers are locked.

3sThe glanceA blurred first screen and its biggest words. “What is this? Would you stay?”
10sThe first screenSharp. They describe it in their own words first, then answer: what is it, who for, what do you get, what does it cost, what does the button do?
1 by 1The scrollOne screen at a time, saying whether they'd keep going. Then what's still unanswered, what convinced them, and whether they'd act.

Keeping them strangers

  • Your brand is hidden. We swap your brand, company and any names you add for invented ones before the visitors see the page.
  • Recognisers are replaced. We ask each visitor, separately, whether it recognises the real brand. If it names it correctly, it's removed and replaced on a different model.
  • No reading ahead. Lower screens aren't sent until the visitor has answered, so first impressions can't borrow from further down.
  • No outside knowledge. Visitors have no web access. Names or numbers in their answers that weren't on the page are flagged and not credited.

Checking against the truth

A score needs a right answer. Yours comes from two places: the one line you give us about what the page sells (the visitors never see it), and a real tap on your main button to see where it leads. We never fill in forms or create accounts.

The rubric: 14 items, 100 points

Most items are the share of visitors who got something right, without hedging. Fixed rules score them, not the visitors' own confidence.

First impression35
Glance names it, first screen names the format, concrete deliverables, who it's for, why this one, what the button does
The offer30
Price and terms on the page, how it works day to day, proof about this offer, no contradictions
Agreement and truth20
Visitors agree after the scroll, and what they agree on matches what you sell
Ready to act15
Questions most visitors are left with, and how many would buy or sign up
80–100
Clear at a glance
60–79
Clear on scroll
40–59
Partly clear
0–39
Unclear
Questions about the score? See the FAQ

Confidence and range

With five visitors, one changed answer can move an item by a few points, so every score comes with a likely range: the score re-calculated with each visitor left out in turn. If the range crosses a band line, we automatically add five more visitors before showing you a result.

Validation

Not yet measuredAgreement with real people on the same pages in a human five-second test. We'll publish the figure here once the first comparison has run, and re-run it every quarter.
Calibration pagesThree reference pages (clear, partial, vague) must score in order with 15+ points between them before any change ships.
VersionedEvery report names its method version. Before-and-after comparisons only use the same version.

What it can't tell you

  • These are simulated visitors. They find the gaps most strangers hit; they don't replace real user testing, recordings or conversion data.
  • “4 of 5 visitors” is not a conversion prediction.
  • Simulated visitors are more patient than real ones, so real drop-off is usually earlier.
  • Names inside images or video can't be hidden. Very well-known brands may be recognised from their products alone; we flag that.
  • Pages behind logins, or that block automated browsers, can't be tested.