Skip to main content

Website Audit Scores Explained

Where each number on your audit comes from: the eight results tabs, the AI recommendations block, and the exact scoring formulas behind them.

Understanding Your Results

Audit results are organized into 8 tabs. Each tab populates progressively as its analyzers complete — skeleton placeholders appear while data is loading.

Overview

Site metadata (title, description, Open Graph tags, favicon), hosting information (server, CDN detection), and the detected technology stack grouped by category.

PageSpeed

Four Lighthouse scores (Performance, Accessibility, Best Practices, SEO), Core Web Vitals (FCP, LCP, TBT, CLS, Speed Index), top optimization opportunities, diagnostics, and a final screenshot of the rendered page.

Accessibility

Two-lens view: User Experience score from Google PageSpeed and a Code Structure analysis covering ARIA usage, semantic HTML, heading hierarchy, form labels, image alt text, and keyboard focus styles. Color contrast is not among them: measuring it needs the rendered page, which a static HTML pass never has. Issues are categorized by severity: Critical, Serious, Moderate, and Minor, and the WCAG line reports which levels those issues sit at rather than claiming a conformance level.

SEO

Composite SEO score (0-100) broken into 8 weighted components: Basic Metadata (20 pts), Link Health (15 pts), Canonical Tags (15 pts), Metadata Consistency (15 pts), Redirect Management (10 pts), Crawling & Indexing (10 pts), Social Media Tags (10 pts), and Schema Markup (5 pts). Each component is expandable with specific recommendations.

AI Readiness

One GEO score (0-100) covering structured data, crawlability, content structure, social tags, readability, alt text, and linking. Outbound links to .gov and .edu domains are counted and shown but score nothing: domain-suffix authority is folklore, not a measured signal. The answer-engine signals below it (FAQ and HowTo markup, question-answer heading patterns, the first paragraph, tables and lists, entity linking) are a checklist with no score of their own, because we have nothing citable to weight them by. Also checks heading hierarchy, reading level, word count, and robots.txt access for 22 AI crawlers — search indexers (OAI-SearchBot, PerplexityBot, Applebot, DuckAssistBot), assistant fetchers (ChatGPT-User, Claude-User, Amzn-User), and training crawlers (GPTBot, ClaudeBot, Amazonbot). Training crawlers are reported but never scored — blocking them is a content-policy choice, not a visibility one. Six specific bots move the score: OAI-SearchBot, PerplexityBot, Claude-SearchBot, ChatGPT-User, Claude-User, and Perplexity-User. Every other crawler on the list, search indexers included, is checked and reported but does not feed the score.

Security

Score gauge based on 6 security headers: Content-Security-Policy, Strict-Transport-Security, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, and Permissions-Policy. Each header shows its status, risk level, and a recommendation if missing or misconfigured. X-XSS-Protection is not among them: modern browsers ignore it, OWASP recommends disabling it outright, and we do not score a header nobody should send.

CSS

CSS quality score with file metrics (count, total size, minification status), complexity analysis (specificity, nesting depth, !important usage), issues by severity, detected frameworks (Bootstrap, Tailwind, etc.), and design token extraction (colors, fonts, spacing, breakpoints).

Hosting

Full SSL certificate details including issuer, validity dates, and days remaining, color-coded so an approaching expiry stands out, plus a DNS records table (A, AAAA, MX, TXT, NS, CNAME, SOA) and server information.

AI Recommendations

Above the results tabs, an AI Recommendations card summarizes the audit in plain English and lists up to 15 prioritized recommendations. Each one shows a severity level, a suggested fix, and a link to the results tab it came from. The recommendations are generated by an AI model from your audit's findings and are constrained to reference only what the audit actually detected — they don't introduce outside data. If a summary can't be generated for a particular audit, the rest of your results are unaffected. Because the recommendations are AI-generated, review them before acting on them or including them in client deliverables. Available on all plans; the block is also included in PDF reports and, on the Agency plan, JSON exports.

Scores & Methodology

Overall Score

The overall score is the average of four Google PageSpeed Insights scores: (Performance + Accessibility + Best Practices + SEO) / 4. This requires the PageSpeed phase to complete. If PageSpeed fails or times out, the overall score is not available.

Color Thresholds

All scores follow the Google PageSpeed Insights color convention:

90 - 100Excellent
50 - 89Needs Improvement
0 - 49Poor

Composite SEO Score (100 Points)

The SEO tab displays a composite score built from 8 weighted components. Each component draws from one or more analyzer results.

ComponentMax PointsHow It's Calculated
Basic Metadata
20
Title presence and length (optimal 50-60 chars), meta description presence and length (optimal 150-160 chars)
Link Health
15
Scaled from broken links analyzer score (score x 15/100)
Canonical Tags
15
Scaled from canonical tags analyzer score (score x 15/100)
Metadata Consistency
15
Scaled from metadata consistency analyzer score (score x 15/100)
Redirect Management
10
Scaled from redirect check analyzer score (score x 10/100)
Crawling & Indexing
10
Sitemap presence (-5 if missing, -1 to -3 per issue), robots.txt presence (-2 if missing, -1 to -3 per error)
Social Media Tags
10
OG Title (-3), OG Description (-3), OG Image (-2), OG URL/Type (-2 if both missing)
Schema Markup
5
0 if no JSON-LD detected, 5 if present

Individual Analyzer Scores

Each analyzer produces its own 0-100 score with specific deduction rules. The analyzer reference gives the exact formula for every one of them.

Next Steps