How Westmark Scan works

What we analyse

Each audit runs Google Lighthouse against the single web page you submit, twice: once emulating a mid-range mobile phone on a throttled connection, and once as a desktop browser. Lighthouse reports four categories:

• Performance — how quickly the page loads and becomes usable (lab measurements such as Largest Contentful Paint, Total Blocking Time and Cumulative Layout Shift).
• Accessibility — automated checks for issues like low colour contrast, missing image descriptions and unlabelled buttons.
• Best Practices — HTTPS, browser console errors, deprecated code and similar quality checks.
• SEO — a technical checklist of basics such as page titles, meta descriptions and crawlability. It is not a ranking audit.

Where the data comes from

We use Google's official PageSpeed Insights API (version 5), which runs Lighthouse on Google's infrastructure and returns the results. We store the category scores, key metrics and the failed or weak checks — we never invent, adjust or randomise scores. If a category or metric isn't returned, the report says “Not measured” instead of showing zero.

These are lab measurements from a controlled test, which makes them consistent and repeatable. They are not the same as real-visitor (field) data, and results can vary a few points between runs.

What the scores mean

Lighthouse scores run from 0 to 100: 90–100 is good, 50–89 needs improvement, and 0–49 is poor.

The Westmark Website Health Score is our own single number for quick comparison. It is a weighted average of the mobile category scores: Performance 35%, Accessibility 25%, Technical SEO 25%, Best Practices 15%. We use mobile because it is the more demanding test and many visitors browse on phones. If any category is missing, we show the analysis as incomplete rather than calculating a misleading score. It is not an official Google score, a ranking, or a prediction of enquiries or revenue.

How AI recommendations are generated

When AI is enabled, we send the structured audit results — scores, metrics and the failed Lighthouse checks — to Anthropic's Claude model. It is instructed to explain the measured findings in plain English, cite the evidence for every recommendation, and never invent numbers. We check its response against a strict format before showing it, and discard recommendations that reference evidence we didn't provide.

The AI does not see your page's design or read its content, so it can't judge your branding, wording or photos. If AI is unavailable, the report uses fixed rule-based explanations and is clearly labelled as such.

Limitations of automated audits

• Only the submitted page is audited — not your whole website.
• Automated accessibility checks catch only some issues; a manual review is needed for full compliance.
• Lighthouse SEO checks technical basics. Rankings depend on content, relevance, reputation and competition.
• Pages behind logins, or sites that block automated browsers, can't be audited.
• Results are cached for up to 24 hours so repeated requests don't waste resources.

How to improve your website

Start with the top three priorities in your report — they're ordered by measured impact. Share the report link or PDF with your web developer, re-run the audit after changes, and audit your other key pages (services, contact, quote request). If you'd like a hand, Westmark Digital can help.