Three use cases. One scoring engine.
Every card below is a live score sitting in the database — click any one to see the per-metric breakdown, judge rationale, and ranked suggestions. To score your own URL, ticket, or PR, you'll need an account.
Score any webpage through a customer lens.
Same judge, four real homepages. The rationale cites copy from each page — these are real scores, not screenshots.
Judge homepage (this site)
Judge.tools leads with a strong, concrete value proposition and backs it with unusually honest trust mechanics (published noise floor, deterministic deltas, versioned judges, tran…
Stripe
Stripe's homepage delivers strong trust signals through quantified statistics and named enterprise case studies, and the dual CTA pattern with Google sign-up reduces signup fricti…
Linear
Linear's marketing page is clearly built for a technical-insider audience: it embeds shell commands, raw code diffs, internal ticket IDs, and unexplained acronyms (PRDs, MCP, Grap…
Vercel
Vercel's homepage is technically dense and clearly written for developers who already know terms like 'Fluid Compute,' 'multi-tenant,' and 'Framework-Defined Infrastructure.' A no…
Score any text artifact — even a Linear ticket.
The spec-completeness rubric scores a vague 'Fix slow dashboard' ticket against scope, acceptance criteria, success metric, and edge cases. Same engine works for any text artifact.
Score a real GitHub PR — diff, description, and all.
A merged Vercel/SWR PR fetched live via the GitHub API and judged by code-quality. Connect the GitHub App and this happens automatically on every PR.