SiteRoastHow SiteRoast scores a landing page
Short version: SiteRoast captures your page on desktop and mobile, then runs three AI review passes (screenshots, visible copy and HTML) over 18 core dimensions, plus programmatic SEO, legal-link and speed checks. Each finding gets a status and an impact weight; those roll up into six pillar scores, and the overall score is the average of the six.
1. What gets captured
We load your public URL in a real browser and take stitched screenshots in desktop and mobile layouts, the visible text, and the page's HTML. We don't install anything on your site and never see your analytics.
2. Three review passes, 18 core dimensions
| Pass | Looks at | Dimensions |
|---|---|---|
| Visual | Desktop and mobile screenshots | Visual hierarchy and layout · Navigation clarity · Readability (font size, line height, contrast) · Scroll experience · Aesthetics and image quality (logo visibility, stock-image authenticity) · CTA visibility · Trust signals · Mobile layout |
| Copy | Visible text | Headline impact · Value proposition (specific or generic?) · Persuasion and tone (authority, specificity) · Objection handling (price, risk, effort, time) · Lead capture (is it clear what happens after you submit?) · One page, one goal (competing CTAs) |
| Technical | HTML | Page speed indicators (heavy scripts, image optimisation, caching hints) · SEO tags (H1 structure, meta description) · Legal compliance (privacy, terms, cookie links) · Mobile form usability (correct input types) |
3. Programmatic checks
- On-page SEO: title present and 30–60 characters, meta description present and 70–160 characters, exactly one H1, canonical, robots meta, Open Graph and Twitter tags, image alt-text coverage, internal and external link counts.
- Legal links: privacy, terms and cookie-policy links detected in the HTML. If all three are present, the legal item is marked Good automatically.
- Speed: Google PageSpeed Insights (Lighthouse) on mobile and desktop where available: performance score, LCP, INP, CLS and TBT, plus the top opportunities.
- Stack detection: common analytics and chat tools (GA, GTM, Meta Pixel, Microsoft Clarity, Hotjar, Intercom, HubSpot and others), so the report can tell you what you can already measure.
4. How the score is calculated
Each finding gets a status and an impact level:
| Status | Points |
|---|---|
| Excellent | 95 |
| Good | 80 |
| Satisfactory | 60 |
| Needs improvement | 35 |
| Failed | 5 |
| Impact | Weight |
|---|---|
| High | 1.5× |
| Medium | 1.0× |
| Low | 0.5× |
Findings roll up into six pillars (UX, conversion, copy, visuals, trust, speed) as impact-weighted averages. The overall site score is the mean of the six pillars. Quick wins are the four worst findings once status and impact are combined (legal compliance is left out of that list).
5. What's illustrative
The revenue-leak range uses monthly sessions × an extra 1%, 2% or 3.5% conversion × average order value × 12 months. Unless the price is read from your page, the order value is a default, and the traffic figure is an estimate, not your analytics. Treat the range as a way to size the opportunity, not a forecast.
6. Limits
- We only see what a logged-out visitor sees on the URL you submit.
- AI review can be wrong. Every finding shows its evidence so you can check it.
- Once you have the traffic, test changes with real visitors; the audit tells you what to test first.