PaperTrail

Prove every number in your citation

Claude reads the literature; a deterministic engine proves every number. Paste an efficacy claim and PaperTrail traces it to the primary source, extracts the actual finding, and flags exactly where the two diverge.

Or try it right now — paste a claim below.

0/2000
Try:Recent checks →
  1. 1Retrieve primary source

    Semantic search over cached PubMed & ClinicalTrials.gov records.

  2. 2Extract the finding

    Structured effect size, population, endpoint, and caveats.

  3. 3Verify & ground every quote

    Compare claim vs source and map each span to the raw text.

One pipeline, from a single claim to a defensible answer

Retrieval, extraction, verification, and synthesis — each step cited, each number checked by rule.

  • Claim verification

    Trace a clinical-trial efficacy claim to its primary source, extract the actual finding, and flag exaggeration or dropped caveats — with a deterministic numeric check on every reported figure.

  • Evidence synthesis

    Pool findings across studies with a random-effects meta-analysis engine, surfacing effect sizes, heterogeneity, and cross-source agreement instead of a single cherry-picked trial.

  • Biomedical evidence

    Ground claims in genetics, drug safety, and target biology — variant pathogenicity, pharmacovigilance signals, and target-disease associations, normalized to canonical entities.

  • Research copilot

    Ask questions across your cached corpus and get answers anchored to exact source spans — a paper-QA assistant that never asserts a fact it cannot cite.

  • Systematic review

    Run PRISMA-style screening and living-evidence tracking, so a review stays current as new trials land rather than freezing on the day it was written.

  • Provenance & audit

    Every flag maps to a verbatim substring of the cached source, and every org action is written to an append-only audit trail you can export.