EVIDENCE & METHODOLOGY

A score you can understand.

See how the numbers are calculated, what supports each finding, and why a source check is only one part of a website review.

1. Collect observable evidence

We request public pages and inspect their HTML and HTTP response headers. Checks cover signals such as title length, alt-attribute presence and HTTPS. We record the page URL, the check result and an explanation of what was observed.

2. Calculate each category

For a category, we count passed checks across the inspected pages and divide by the total number of checks in that category. We multiply by 100 and round to the nearest whole number. Checks within a category have equal weight.

Illustrative calculation

If 4 of 6 checks pass in a category, its score is round(4 ÷ 6 × 100) = 67. This example explains the arithmetic; it is not a result from a customer website.

3. Calculate the overall score

We take the average of the seven rounded category scores, then round that average to a whole number. Every category contributes equally, even when categories contain different numbers of checks. The overall score is not the pass rate of all 17 checks combined.

4. Organise the improvement plan

Checks that do not pass become review findings with an action, evidence and a preset High, Medium or Low priority. Priority is an ordering aid built into the check; it is not a prediction of lost revenue, business impact or exploitability.

The thresholds are review rules

For example, the current title check accepts 10–65 characters and the meta-description check accepts 50–170. The content check looks for at least 150 approximate words. These are simple tool thresholds, not Google requirements. A useful page can fall outside them. An expert review should consider purpose, audience and context.

What the score does not measure

Google rankings, traffic, enquiries or sales.
Rendered layout, contrast, keyboard use or full accessibility compliance.
Real-user speed, Core Web Vitals or Lighthouse performance.
Malware, vulnerabilities, penetration-test results or legal compliance.
The quality or accuracy of writing, alt text or security policies.

Coverage and uncertainty matter

The scan is limited to up to five pages, with response-size and timeout limits. Some sites block automated access. A site that cannot be scanned receives an error rather than an invented score. Later pages that fail may be omitted with a warning. Check the listed pages and warnings before treating a report as representative of the whole website.

AI assessments and PageSpeed integration are not active in the early-access version. The findings are generated from the defined source checks, not an AI judgement.