How we score a landing page

Every PageCritic score is the sum of 36 checks with fixed points, grouped in 7 areas that add up to 100. The AI never picks the number. 25 checks (62 points) are measured in code from the page itself, in a real browser on a desktop and a phone screen. The other 11 (38 points) need judgment, so the AI answers each one by picking a level from the written descriptions below, quoting the page first; it is asked five times and the middle answer counts. Every check that loses points becomes one issue in the report, ranked by the points a fix wins back.

The areas and their weights follow what conversion research agrees on: a visitor who can’t tell what a page offers never reaches the button, so clarity and the value proposition weigh most, and speed and search basics least.

Score	Band	What it means
90–100	Excellent	Little left to fix on the page itself.
75–89	Strong	Well built; the gains are in testing.
60–74	Decent	Works, with clear fixes left.
40–59	Needs work	The basics are there; it leaks most visitors.
0–39	Weak	Visitors can’t tell what it is or what to do.

Clarity: 20 points

A visitor who can't tell what this is in a few seconds leaves before reading anything else. MECLABS and LIFT both rank it first.

0: nobody can tell what it is or who it's for · 5: you can guess the area, not the product or the audience · 12: what it is and who it's for are clear; the headline is a slogan · 20: a stranger can say what it is, who it's for and what they get after one look

Check	Points	How it’s scored
Says what it is C1 · Judged	5	First screen only (screenshot 1): could a stranger say what kind of product or service this is, its category? 5 · the first screen names or shows the category plainly 2 · you can guess the area, not what it is exactly 0 · no way to tell from the first screen
Says who it's for C2 · Judged	4	First screen only: does it name who it's for (a role, a situation, an industry, a kind of customer), or make it unmistakable? 4 · a specific audience is named or unmistakable 2 · only a generic group: teams, businesses, everyone, creators 0 · no audience at all
Headline states a result C3 · Judged	4	First screen: what does the main headline (the largest text) say? 4 · names a change for the visitor: something they get, save, achieve or stop suffering ("Get paid twice as fast", "Ship your backend in a weekend") 2 · names the product, its category or a capability, without that change ("Privacy-friendly Google Analytics alternative", "The CRM for agencies") 0 · a slogan, brand line, question or vague claim ("Work, reimagined", "Built for the future")
An AI reads it the same way C4 · Measured	4	A separate model reads only the page text and writes a one-line summary; a second compares it with the headline. match 4, partly 2, mismatch 0.
Headline is short and prominent C5 · Measured	3	In the desktop first screen: the headline is the biggest text (1), 2–12 words, two lines or fewer (1), at least 28 px (1).

Value proposition: 15 points

Why want it, and why this one: the core of MECLABS' 3v and the centre of the LIFT model.

0: features and adjectives only · 5: benefits stated in general words, no reason to pick it · 10: a specific benefit, and the offer is clear · 15: specific, set against the alternative, offer and price clear

Check	Points	How it’s scored
Specific benefit V1 · Judged	4	Whole page (read the page text, not only the screenshots): is there a concrete, checkable benefit, such as a number, a timeframe or a named result? 4 · at least one benefit for the visitor with a number, a timeframe or a named result ("set up in 5 minutes", "cut support tickets by 30%", "get your first sale this week"); customer counts are proof, not benefits 2 · benefits only in general words ("save time", "grow faster", "easier") 0 · only features or adjectives
Why this one V2 · Judged	4	Whole page: does it say why this instead of the alternative (a named competitor, the manual way, or doing nothing)? 4 · names or plainly refers to an alternative (a competitor, "spreadsheets", "agencies", "doing it by hand") and says how this differs 2 · a distinct angle with nothing named to compare it to ("the only…", "the simplest…", "privacy-first") 0 · nothing sets it apart
Benefits, not just features V3 · Judged	3	Take the first three section headings after the hero (the h2 list in the page facts), skipping bare labels like "Features", "Pricing" or "FAQ". How many state what the visitor gets (a result), not only what the product has? With fewer than three, the missing ones count as no. 3 · at least two of the three 1 · one of the three 0 · none
Price or "free" on the page V4 · Measured	2	A price, "free", "trial" or a pricing link is on the page.
Offer is clear V5 · Judged	2	Whole page, before clicking anything: does a visitor know what they get and what it costs, or that it's free? 2 · what you get and the price (or "free") are both on the page 1 · one of the two is missing or vague 0 · neither

Call to action: 15 points

The one place a visit turns into a sign-up or a sale (Fogg's prompt).

0: no clear next step · 5: a button exists but is buried, generic or competes with others · 10: clear and in the first screen, with a few rivals · 15: one obvious, specific action, in the first screen on both screens and repeated below

Check	Points	How it’s scored
Button in the first screen A1 · Measured	4	A button (or a link styled as one) with an action label in the first screen on desktop (2) and on a phone (2).
One main action A2 · Measured	3	Distinct button labels in the desktop first screen: up to 2 earns 3, 3 earns 2, 4 earns 1, 5 or more 0.
Button says what you get A3 · Judged	3	First screen: read the label of the main call-to-action button (the most prominent one). 3 · names what they get or the next step: a noun like trial, demo, audit, report, quote, account, template, or a possessive ("Start free trial", "Book a demo", "Get my audit", "Create free account") 1 · a verb with no object: "Get started", "Sign up", "Start now", "Try it", "Join", "Contact us" 0 · "Learn more", "Submit", "Click here", "Read more", or no button
Button stands out A4 · Measured	3	The main button's colour has at least 3:1 contrast with what's behind it (WCAG 1.4.11) (2) and it is at least 40 px tall (1).
Action repeated down the page A5 · Measured	2	The main button (same label or link) appears again below the first screen. Pages under 1.5 screens long pass.

Trust: 15 points

Anxiety stops people who already want it (MECLABS −2a; Stanford Web Credibility guidelines).

0: anonymous, unproven, looks unfinished · 5: a real company, but no proof · 10: proof, but not where the decision happens · 15: named proof next to the button, the risk of trying removed

Check	Points	How it’s scored
Social proof T1 · Judged	4	Whole page (read the page text, not only the screenshots): what proof that other people use and like it does the page show? Customer or company names, "used by" lists and user counts in the text count. 4 · named testimonials (name with role, company or photo), recognizable customer logos, ratings with their source, or concrete usage numbers 2 · unsourced claims ("loved by thousands") or anonymous quotes 0 · no proof
Proof where the decision is T2 · Judged	2	First screen: is any proof (a testimonial, logo, rating or usage number) visible in it, or right beside the main button? 2 · yes 0 · no, only further down, or none
A real company behind it T3 · Measured	3	Links to contact, about, privacy and terms: all 4 earn 3, 3 earn 2, 2 earn 1.
Risk of trying removed T4 · Judged	3	Whole page: does it lower the risk of trying it (a guarantee, a free trial or free plan, no card needed, cancel anytime, refunds, or a clear privacy or security reassurance near the offer)? 3 · at least one of these is stated 0 · none of these
Nothing looks unfinished T5 · Measured	3	Starts at 3: a visible site-builder badge −2, placeholder text (lorem ipsum, "Your company") −2, a broken image in the first two screens −1.

Focus and design: 10 points

Every extra exit, field or pop-up takes attention from the one action (attention ratio, MECLABS friction).

0: cluttered, covered by pop-ups, long forms · 4: one focal point, but many exits or a long form · 7: focused, a few distractions · 10: one focal point, few exits, nothing in the way

Check	Points	How it’s scored
Few exits in the first screen F1 · Measured	3	Clickable links and buttons in the desktop first screen, menu included: up to 8 earns 3, up to 12 earns 2, up to 18 earns 1.
Short forms F2 · Measured	2	The longest form asks for up to 3 fields (2), 4–6 (1), 7 or more (0). No form passes.
Nothing covers the first screen F3 · Measured	2	Pop-ups, cookie banners and other pinned boxes (not the top bar) cover up to 10% of the first screen (2), up to 25% (1), more (0).
Easy to scan F4 · Measured	3	Section headings: at least one per 1.7 screens of page (2), or one per 3.3 screens (1); short pages with a heading pass. At most 2 paragraphs longer than about 5 lines on desktop (1).

Mobile and accessibility: 15 points

Most landing-page visits come from phones; these checks follow WCAG 2.2 and Lighthouse.

0: breaks on a phone · 5: usable on a phone with effort · 10: works on a phone, some readability gaps · 15: fits, readable, easy to tap, passes the contrast checks

Check	Points	How it’s scored
Fits a phone screen M1 · Measured	3	Viewport tag present and zoom not blocked (1); no sideways scrolling at 390 px wide (2).
Headline in the first phone screen M2 · Measured	3	On a phone the whole headline is in the first screen (2), starting in its top 60% (1).
Readable text size M3 · Measured	2	On a phone, body text is at least 14 px (1) and less than 40% of the text is under 12 px (1, the old Lighthouse rule).
Easy to tap M4 · Measured	2	Buttons and links smaller than 24×24 px with a neighbour too close (WCAG 2.5.8): up to 5% of targets (2), up to 20% (1).
Text contrast M5 · Measured	3	Text in the first screen below 4.5:1 contrast (3:1 for large text, WCAG 1.4.3): none (3), 1–2 (2), 3–5 (1), more (0). Text over images is skipped.
Alt text and language M6 · Measured	2	Content images have alt text (1); the page declares its language (1).

Speed and SEO: 10 points

Hygiene: rarely why a page fails to convert, but slow pages and bad search snippets cost visits before anyone reads.

0: slow, unshared, unfindable · 5: loads, but heavy or missing search basics · 10: fast, light, with a title, description and share preview

Check	Points	How it’s scored
Loads fast and light S1 · Measured	3	Page weight over the wire up to 2.5 MB (2) or 5 MB (1); largest content painted within 2.4 s on desktop, measured from our server (Lighthouse's desktop median) (1).
Layout doesn't jump S2 · Measured	2	Cumulative layout shift up to 0.1 (2), up to 0.25 (1) (Core Web Vitals limits, lab measurement).
Search title S3 · Measured	2	A page title (1) that says more than the brand: 3+ words and up to 70 characters (1).
Search description S4 · Measured	1	A meta description of 50–160 characters, so the search snippet is yours.
Share preview S5 · Measured	1	An og:image and og:title, so links shared in chats and social posts show a card.
Can be found S6 · Measured	1	Served over HTTPS and not marked noindex.

What this can’t measure

Speed is a lab measurement from our server, unthrottled. It flags heavy pages; it is not your visitors’ Core Web Vitals.
Text over images and gradients is left out of the contrast check, since its background can’t be read from the styles.
We don’t know where your traffic comes from, so ad-to-page message match isn’t scored. Neither is your real conversion rate: the score says how ready the page is, not how it performs.
Judged checks still rest on an AI’s reading. Fixed levels, quoted evidence and five samples make it repeatable; they don’t make it infallible.

Sources

Rubric version 2, in use since 24 October 2026. Reports made before that used an older method, where the AI chose the scores, and their numbers don’t compare with these.

Audit my page free