What Claims is
Claims produces a machine-readable version of the scientific record that is optimised for AI models and agentic research workflows. It reads scholarly publications and turns what they claim into structured, machine-optimised records: each claim is linked to its source evidence, the strength of the evidence is scored, and ontology mappings make connections across papers visible.
Independent research supports this approach: a 2026 computer-science study found that turning papers into structured, evidence-grounded research records raised AI agents’ question-answering accuracy from 72% to 94% (Liu et al., 2026).
We created an open R&D competition for Claims that rewards people for helping us to continuously improve the quality of this data. Learn how you can join us here.
As a first step, we will produce this data on demand for biotech companies.
Recognition
Claims won Proof of Pitch 2026 at the Louvre in Paris, June 2026.
Documents
Claims white paper: Selective verification, continuous learning, and identity-robust incentives. P.D. Koellinger, C. Roessler, O. Ugot. Version 1, August 2026. claims111.ai/whitepaper.
The game theory behind Claims: Verification without known answers – The economics of reusable judgment. C. Roessler. Preprint, September 2026.
claims111.ai/papers/verification-without-known-answers