Website Audit
Available on: Power plans and higher (within the Performance Hub)
The Website Audit checks your brand’s site against the Website Specification, an open standard of technical best practices for the modern web. It scores your homepage and site-root resources across foundations, SEO, security, agent readiness, performance, and resilience, then turns every failure into an actionable fix your AI agent can carry out.
The audit runs automatically when you add a brand, on a weekly schedule, and on demand. Its score contributes 10% of your Brand Visibility Score.
What It Checks
The MVP audit runs roughly 45 automated checks grouped into six categories. Each check maps to an item in the Website Specification and carries a severity based on whether the spec marks it as required, recommended, or optional.
| Category | Examples | How it is checked |
|---|---|---|
| Foundations | doctype, lang, charset, viewport, title, meta description, canonical, favicon, Open Graph | Homepage HTML parse |
| SEO | robots.txt, sitemap.xml, meta robots / X-Robots-Tag, single H1, JSON-LD structured data | HTML + /robots.txt, /sitemap.xml |
| Security | HTTPS redirect, HSTS, CSP, X-Content-Type-Options, frame protection, Referrer-Policy | Response headers |
| Agent Readiness | /llms.txt, robots AI user-agents, structured data, stable homepage URL | HTML + robots parse |
| Performance | compression (content-encoding), cache-control on HTML, image dimensions | Headers + HTML heuristics |
| Resilience | custom 404 behavior, /.well-known/security.txt | Well-known and synthetic URL probes |
Checks that cannot be evaluated automatically are marked skipped and never count against your score.
How Scoring Works
Each automated check is weighted by its spec status:
- Required: weight 2
- Recommended: weight 1
- Optional: weight 0.5
- Avoid (inverted): weight 2 - passing means the anti-pattern is absent
A passing check earns full credit, a warning earns partial credit, and a failure earns none. The overall score (0-100) is the weighted pass rate across all automated checks, and each category gets its own sub-score. Skipped and errored checks are excluded from the math.
The latest completed audit (if no more than 14 days old) feeds the Website Audit pillar of the Brand Visibility Score at a 10% weight.
When It Runs
| Trigger | When | Notes |
|---|---|---|
| Onboarding | When you add a brand | Runs in the background; progress shows in the wizard |
| Manual | ”Run audit” button in the Website Audit tab | Rate limited to once per hour per site |
| Weekly | Every Monday 6:30 AM UTC | Runs before the weekly performance digest |
Eligible sites are active brands with a domain on a plan that includes the Performance Hub.
The Website Audit Tab
Find it under Performance → Website Audit.
- Score summary: Overall score, change since the last run, and pass/fail/warn/skip counts.
- Category breakdown: Sub-scores for each of the six categories.
- Checklist: Every check with its status, severity, message, and evidence (the exact header value or HTML snippet observed). Filter by category or status.
- Opportunities: Prioritized, agent-actionable fixes.
Fixing Issues with Your Agent
Each opportunity includes:
- A plain-language fix hint and a link to the relevant spec section.
- Fix with Agent: opens Cuppa Chat pre-filled with a ready-to-run instruction.
- Copy for MCP: copies a JSON payload for use with an external agent.
- Resolve / Dismiss to manage the opportunity once handled.
Notifications
The weekly audit can alert you when your score moves materially:
| Event | Trigger | Default channels |
|---|---|---|
performance.audit_regression | Score dropped 5+ points or several required checks newly failing | Email + Slack |
performance.audit_improvement | Score gained 5+ points | Off by default |
The weekly performance digest also includes your audit score and its week-over-week delta. Manage these under Team Settings → Notifications.
API, MCP, and CLI Access
REST API
| Endpoint | Method | Description |
|---|---|---|
/v1/performance/website-audit | GET | Latest audit summary + category scores |
/v1/performance/website-audit | POST | Enqueue a fresh audit (1/hour/site) |
/v1/performance/website-audit/history | GET | Score history over time |
/v1/performance/website-audit/checks | GET | Per-check results (?audit_id= optional) |
/v1/performance/website-audit/opportunities | GET | Open opportunities |
/v1/performance/website-audit/opportunities | PATCH | Resolve / dismiss an opportunity |
All endpoints take a site_id query parameter and require a Power plan or higher.
MCP / CLI Tools
| MCP tool | CLI command | What it does |
|---|---|---|
get_website_audit | cuppa website-audit overview | Latest score, deltas, category scores |
get_website_audit_opportunities | cuppa website-audit opportunities | Open, actionable fixes |
run_website_audit | cuppa website-audit run | Enqueue a fresh audit |
cuppa website-audit overview --site-id <id>
cuppa website-audit opportunities --site-id <id>
cuppa website-audit run --site-id <id>The same three tools are available to in-app agents in Cuppa Chat, so you can ask your agent to “audit my website and fix the critical issues.”
Troubleshooting
”Audit failed”
The audit fetches your live homepage and root files. If the site is unreachable, behind a WAF that blocks automated requests, or returning errors, the run is marked failed with the reason shown in the tab. Retry once the site is reachable.
”Score seems low”
Open the checklist and sort by failing required checks first. Required checks carry the most weight, so fixing those moves the score fastest. Use the Opportunities panel to hand the fixes to your agent.
”Skipped checks”
Some spec items cannot be verified automatically (for example, anything requiring human judgment). These are marked skipped and never lower your score.
Related Docs
- Performance Hub - the dashboard the audit lives in
- Agent Brand Engineering - use audit data with AI agents
- Website Specification - the open standard the audit is based on