Methodology

The open Web Health Score methodology.

Five pillars, twenty points each, one composite score out of 100. Scored against the same standards a top-tier UK digital agency would cite in a £18-35k audit. Every weighting and rubric is published below. You can audit our audit.

Last published version: v1.0 (May 2026). The methodology is versioned. Older audits remain valid against the version they were scored under.

Score bands

What the number means.

The composite score is computed by Brief from the six pillar scores (each scored out of 20) and presented as a single number out of 100. The band names below are the same names used in every audit Wingenious has run since April 2026.

Leading 90–100

Category-defining. Sets standards others follow.
On standard
Strong 75–89

Few material gaps. Polished and trustworthy.
On standard
Baseline 60–74

Meets the contemporary professional standard for a 2026 UK commercial website.
At standard
Lagging 45–59

Functional, but below the contemporary standard in several measurable respects.
Below standard
Sub-standard 30–44

Multiple simultaneous findings below the contemporary standard.
Critical
Crisis 15–29

Fails one or more basic expectations of a 2026 UK commercial website.
Critical
Inoperable 0–14

Fundamentally broken across multiple dimensions.
Critical

The six pillars

What we measure. How we measure it. What we cite.

Pillar 01

Agent: Scout
Weight: /20

Findability

Whether the right people can discover you on Google, Bing, and the answers given by ChatGPT, Perplexity, and Claude.

What we measure

Indexability (robots.txt, canonical, sitemap, noindex audit)
Title and meta description quality across crawled pages
Open Graph and Twitter card completeness
JSON-LD structured data per page type (Organization, LocalBusiness, Product, Article, FAQ)
Internal linking depth and orphaned-page detection
AI Visibility: llms.txt presence, robots.txt directives for GPTBot / Claude-Web / PerplexityBot, sitemap signals

Standards cited

Google Search Central documentation
schema.org core types
BS EN 16234-1:2019 (UK e-Competence)
Bing Webmaster Guidelines

Pillar 02

Agent: Sprint
Weight: /20

Performance

How fast, stable, and resilient the experience is on a real customer’s 4G phone in the back of a taxi.

What we measure

Lighthouse Core Web Vitals on mobile (LCP, INP, CLS, TBT)
Total page weight and render-blocking resources
Image format efficiency (AVIF / WebP / fallbacks)
JavaScript payload audit per page
Third-party script inventory and execution cost
Hosting transport (HTTP/2, HTTP/3, TLS 1.3, compression)

Standards cited

Google Core Web Vitals thresholds (mobile)
web.dev performance guidance
IETF HTTP/3 (RFC 9114), TLS 1.3 (RFC 8446)

Pillar 03

Agent: Anchor
Weight: /20

Credibility

Whether the site convinces a sceptical UK SME visitor that you are trustworthy enough to send their first enquiry.

What we measure

Above-the-fold trust signals (logos, badges, accreditations, named clients)
About / team / ownership transparency
Companies House presence and consistency
External reviews (Trustpilot, Google, Reviews.io) surface and consistency
Press / case studies / named customers
Security and privacy posture (response headers, cookie consent, privacy page)

Standards cited

ICO guidance on cookies (July 2023)
UK GDPR, PECR 2003
OWASP Secure Headers Project
Nielsen Norman Group trust research

Pillar 04

Agent: Lens
Weight: /20

Clarity

Whether the message is readable, accessible to WCAG 2.2 AA, and unambiguous in the first fifteen seconds.

What we measure

WCAG 2.2 AA via axe-core run on every crawled page
Reading age (Flesch / Flesch-Kincaid) per page type
Heading hierarchy and landmark structure
Keyboard operability and focus management
Reduced-motion respect and target-size compliance
Five-second test: can a stranger explain what you do?

Standards cited

WCAG 2.2 Level AA
Equality Act 2010 (Public Sector Bodies Accessibility Regulations 2018 informative)
Nielsen 10 usability heuristics
Cognitive Walkthrough method

Pillar 05

Agent: Compass
Weight: /20

Conversion

Whether the journey from landing to enquiry actually works on the device and locale of your customers.

What we measure

Primary CTA presence, specificity, and ranking (no "Learn More" failure mode)
Form completeness, anti-spam, and accessibility
Sticky / repeat CTA placement on long pages
Friction audit on enquiry path (clicks-to-action)
Local intent signals (NAP consistency, GBP, area-serving microdata)
Mobile-first conversion path (one-thumb operable, no desktop reflow)

Standards cited

Baymard Institute B2B CRO benchmarks
CXL Institute conversion research
Nielsen Norman Group five-second testing
Fitts · Hick · Miller · Jakob · Doherty UX laws

Per-finding severity

Every finding carries a severity grade, a cited standard, and a commercial consequence.

We do not surface generic recommendations. Every finding documents what was observed, the standard it is measured against, and what it costs you if you leave the configuration unchanged.

CRITICAL
Regulatory exposure, legal liability, or active revenue leak. Fix this week.
HIGH
Below the contemporary standard. Materially affects conversion or trust.
MEDIUM
Noticeable below-standard configuration. Fix in the next release cycle.
LOW
Polish-grade improvement. Worth doing once everything above is shipped.

What you get at each tier

Same methodology. Different depth.

Depth of audit by tier
Tier	Pages crawled	Deliverable	Turnaround
Free audit	5 pages	Web Health Score + top issue per pillar + PDF	~90 seconds
Starter audit (coming soon)	50 pages	Full prioritised action list, competitor comparison, 25-40 page PDF	Under 30 minutes
Pro audit (coming soon)	Full site	90-day roadmap, GA4 + Search Console connected, accessibility report per WCAG criterion, security headers audit, written 30-50 page document	48 hours

How we choose the pages

Five pages, not five hundred.

A focused sample beats a comprehensive crawl for the questions this audit is built to answer. The five pages we pick are the ones where credibility, conversion, and findability signals actually live.

Step 1 · The URL you submitted

The root page becomes page 1. Most home pages set the brand voice, the trust strip, and the primary call to action.

Step 2 · Conversion keyword pass

We scan every internal link on the home page and pick up to four whose URL path contains one of the SME conversion keywords: about, services, products, contact, pricing, shop, work, case studies, team, company. The first match wins, in the order links appear in the document.

Step 3 · Top-level fallback

If fewer than four keyword pages exist, we top up with same-host links whose paths are one or two levels deep. Blog posts, dated archive pages, and external links are skipped. Editorial nav is favoured over deep content.

Honest gaps in the current selection

The crawler does not yet read sitemap.xml for page selection, only for presence-detection. A future revision will prefer sitemap-listed URLs over keyword scanning.
The keyword list is English-only. A French or German site will fall through to the depth-based fallback in step 3.
We do not currently weight <nav> links higher than footer links. Primary navigation is a stronger editorial signal and will be prioritised in a future revision.
Pages chosen in step 2 are the first four matches in DOM order, not the four best matches by importance.

Each audit report lists the exact URLs analysed and the reason each was picked, so the selection can always be cross-checked against the findings.

About this audit

The audit is free. It is produced by AI (Claude Sonnet 4.6 with adaptive thinking) running the six-pillar methodology documented on this page. The same methodology delivered by hand at a top-tier UK digital agency typically bills at £18k to £35k + VAT. You read it at no cost because it is the shortest demonstration of how we work: AI-accelerated, human signed, delivered in minutes rather than months.

What AI does, what humans still do. AI compresses the costly parts of an audit: crawl, fingerprinting, header inspection, schema extraction, competitor fetches, accessibility testing, performance instrumentation, structured-data validation, copy assessment. Craft still matters. A thirty-year-experienced digital lead (Gary Cheers, Wingenious Ltd) signs every methodology decision off. AI buys back the calendar, not the standard.

Why the audit is free. It is the proof of the offer. If the audit finds something serious, Wingenious can rebuild, remediate, or coach you through fixing it. Most owners run the audit, action what they can themselves, and never speak to us. That is fine. The audit is yours either way.

Versioning

The methodology is versioned, citable, and challengeable.

Current version: v1.0, published May 2026. Material changes (new pillar, change to weightings, change to band thresholds) increment the major version. Wording and rubric clarifications increment the minor.

Older audits remain valid against the version they were scored under. If we change how Findability is measured, your previous score is not retrospectively reduced. Your next audit will be scored against the latest version with the difference noted.

Disagree with a finding? Use the methodology challenge form with the audit ID and the specific finding. We reply with the rubric used, the evidence captured, and either correct the audit or explain the decision.

Full methodology document: available as a downloadable PDF at docs.freewebsiteaudit.ai (Starlight subsite, shipping after launch).

See the methodology in action

Run an audit. Get the evidence.

Every finding cites its standard. Every score traces back to a measurement on this page.

Run my free audit See an example report