About SourceScore

Why this exists

AI engines surface a small subset of sources as authoritative citations. The criteria are opaque: there’s no public list, no published ranking, no transparent rubric. SourceScore publishes a transparent rubric and ranks sources against it. Anyone can re-derive any score from the underlying signals. And on the API side, every verified claim ships with its primary sources, an HMAC-SHA256 signature, and a ready-to-paste citation — built for grounding LLM responses without re-doing source-quality work yourself.

Who we are

SourceScore is operated by a small independent team. The site is a sister project to HoldLens (sec-filings reference index) and other reference-grade fleet sites. The methodology is our intellectual property; the underlying public-source data we score and the verified claims we publish are credited to their original publishers under CC-BY 4.0.

Editorial policy

• No paid placements. Scores cannot be purchased.
• Methodology changes are versioned and announced.
• Every score and every claim has explicit signals you can verify against primary sources.
• Corrections are timestamped and public.
• We do not auto-generate or fabricate data.
• Verified claims at confidence <0.70 are not published. Performance comparisons are intentionally excluded from v0 because benchmark numbers depend on version + prompt format.

Two products, one methodology

SourceScore Index ranks sources by their AI-citation fitness. VERITAS ranks individual claims by their sourcing depth. Same trust-signal thinking, two different units of analysis. Both products are published methodology; both are independently re-derivable.

Get in touch

Questions about methodology, corrections, partnership requests, paid-tier API early-access, or anything else: contact us or email [email protected].