“Every claim cited” — how, exactly
Doctors are right to be skeptical of AI answers. This page is the audit trail: what goes in, how it is ranked, and what we refuse to do.
A curated, license-cleared corpus — not the open web. Guidelines, formularies and peer-reviewed literature are chosen source by source.
Every source carries an authority grade in a fixed, visible hierarchy: Indian national guidance first, observational evidence last.
Answers are generated only from retrieved passages. A claim without a resolvable citation is discarded, not published.
Guidelines get revised — so does the corpus. Sources are versioned and re-ingested when their publishers update them.
Ranked in the open
When sources disagree, rank decides what leads the answer — and the losing positions stay visible with their own citations. The hierarchy is fixed, published, and India-first by design.
All positions, side by side
Disagreement across sources is detected and surfaced — each position shown with its source and rank. The answer never hides a dissenting guideline to look cleaner.
A warning, not a guess
Where evidence is thin, the answer carries an explicit low-confidence banner and shifts source-forward — you read the evidence itself instead of a synthesized summary.
EvidenceFirst supports — never replaces — clinical judgment. If an answer and your clinical assessment disagree, trust your assessment and check the sources.
Ready for the ward round.
EvidenceFirst is live. Create a free account with your work email and ask your first question in under a minute.
For physicians, residents and students · Free during rollout