MethodologyLast updated September 2026
How we score pages and sections
A person writes the criteria. An AI checks each one against the page. The score is the share of criteria met. This page shows how, so you can check a score yourself, or disagree with it.
Show every company how its website compares with its competitors, with evidence.
The trend tracker for websites, category by category.
Every score on Web Anatomy comes from the same method. The criteria were written by Gabriel Amzallag, from conversion copywriting, UX research and the patterns that repeat on pages that convert. An AI applies them the same way to every company, so two scores can be compared.
One section, one checklist, one score.
A hero is not judged like a pricing table. Each type of section has its own checklist, and a section is only scored against the checklist for its type.
- 01Step 01, The criteria
Written by a person, one sentence each
Each criterion says one thing a strong section does, in a sentence you can check. "States who the product is for." "Shows the real product, not an illustration." "Reduces the risk of trying, for example no credit card required."
Next to the criteria sit the rules for what does not count. They are there to stop generous scoring:
- A sign-up button in the navigation bar is not a call to action in the hero.
- An abstract illustration or a stock photo is not a product visual.
- Social media icons are not social proof. Logos, customer counts and ratings are.
- "Easy" or "powerful" is not a competitive claim.
- "Fast" without a number is not a time promise.
- 02Step 02, The check
An AI looks for visible evidence
An AI looks at the screenshot of the section, with the page it sits on as context, and goes through the checklist one criterion at a time. A criterion counts only when there is clear visible evidence for it. When in doubt, it does not count.
For every criterion that passes, the AI quotes what it saw: the button label, the number, the logos. That quote is shown next to the screenshot, so you can check it.
If the screenshot is not a real section, such as a cookie banner, a pop-up or a blank band, it is not scored at all.
- 03Step 03, The score
The share of criteria met
The score is the share of the checklist the section meets, from 0 to 100. Every criterion counts the same. There is no bonus for style and no penalty for taste.
Criteria come in two kinds, and both are scored.
- Features
- Something the section has, that you can point at in the screenshot: a logo wall, a second button, a price per plan.
- Practices
- A judgment on how well the section does something: how specific the benefit is, how clearly it names the audience.
The whole page, category by category.
A page is more than its sections. It is also scored as a whole, against a checklist grouped into categories.
- 01Step 01, The checklist
Pass rules, grouped and weighted
The page checklist is grouped into categories: hero, value proposition, copywriting, trust and credibility, conversion, design and UX, and mobile. Each item has a pass rule that leaves no room for interpretation.
For example: "The hero states both what the product is and who it is for. Missing either one is a fail."
Each item also has a weight. The critical items, the ones a visitor notices in the first seconds, weigh the most.
- 02Step 02, The check
Pass, fail, or not evaluable
An AI reads the text of the page, the desktop screenshot and the mobile screenshot, and gives each item one of three verdicts with a one-line reason.
- Pass
- The evidence clearly and fully meets the rule.
- Fail
- The evidence contradicts the rule, is missing, or only half meets it.
- Not evaluable
- The rule does not apply to this page, or the evidence it needs is missing. It is left out of the score, not counted as a fail.
- 03Step 03, The score
Weighted by category, then averaged
Each category score is the weighted share of its items that pass. The page score is the average of the category scores, so one strong category cannot hide a weak one.
When a page is compared with its competitors section by section, its score is the average of its section scores instead, so every page in the comparison is measured the same way.
A person sets the rules, the AI applies them.
- 01The criteria
Defined and revised by Gabriel
Every checklist and every rule for what does not count was written by Gabriel Amzallag. When the AI tags something it should not, the rule is tightened and the checklist gets a new version, so the same mistake is not made twice.
- 02The evidence
Every verdict can be checked
A score is never shown without its reasons. Each passed criterion comes with the quote the AI saw, and each section comes with its screenshot, so a score can always be checked against the page.
- 03The library
Reviewed before it is published
Pages in the public library are reviewed by a person before they are published. Nobody can pay to appear, buy a better score, or remove a bad one.
Think a score is wrong? Tell us.
If you believe a score is wrong, email gabriel@webanatomy.ai with the URL and the criterion you disagree with. Every dispute is reviewed, and when a rule was wrong, the rule is fixed for every page.
Score your page against your competitors.
The same checklists, applied to your page and to the pages you compete with.