AI system from QED Science ranks top 1% of bioRxiv preprints, sparking debate

QED Science, a private firm headquartered in Tel Aviv, has developed an AI-driven platform that examines the validity and originality of life-science manuscripts by comparing reported claims with supporting data and identifying missing evidence. The system, offered free to researchers, was trialed on the preprint server bioRxiv after a partnership announced by openRxiv last November. In a June analysis, the AI ranked more than 57,000 preprints posted from May 2025 to April 2026, isolating the highest-scoring 574 papers as the top 1% based solely on the tool's internal scores.

The company claims the ranking uncovers “hidden gems” that may have been overlooked by traditional publishing routes, citing a validation study where its scores outperformed journal rankings in 12.9% of cases. Nonetheless, many scientists warn that the list could reinforce metric-driven prestige and doubt the ability of AI to evaluate research quality transparently. QED Science emphasizes that the platform runs without human editorial input but continuously incorporates feedback from its user community to improve performance.