# Extractability Oracle

> Pay-per-call schema/JSON-LD extractability scoring (0-100) for any web page: how well AI
> crawlers and answer engines can understand and quote it (GEO). Signed responses, x402 on Base.

## Endpoints
- GET /extractability?url=<url> — paid $0.01 (USDC on Base via x402), signed JSON: score + breakdown + fixes
- POST /extractability — paid, score raw HTML posted as the body
- GET /extractability/preview?url=<url> — free, unsigned, rate-limited
- GET /health — liveness
- GET /provable/head — hash-chain head + count
- GET /provable/verify — re-hash and verify the whole scan chain
- GET /.well-known/x402 — machine-readable manifest

## Rubric v2.0 (0-100)
Structured data / JSON-LD (25); Extractable facts & patterns (20); Metadata (15);
Structure & headings (15); Machine readability / AI access (15); Authority & entity (10).
Deterministic: same input -> same score. Paid output adds per-check subscores + prioritized fixes.

## Proof
Every paid scan is appended to a tamper-evident sha256 hash chain; walk it from /provable/head and
check each link with /provable/verify. Every paid response carries signed_by + signature.