AI Tool Rates Top 1% Preprints: Should Scientists Trust It?

A new AI tool created by QED Science, a startup based in Tel Aviv, Israel, is designed to evaluate life sciences research for originality and validity. The tool assesses whether the claims in a manuscript are supported by the data and identifies any gaps in the work. Over 10,000 labs across 1,500 institutions in more than 70 countries have used it for free.

In November, bioRxiv and medRxiv began testing this tool. QED Science analyzed over 57,000 preprints from May 2025 to April 2026 and ranked them, selecting the top 1%. However, this ranking has caused debate. Some worry it could add another layer of prestige-based metrics in academia, while others question AI’s ability to fairly judge research quality.

QED Science’s tool works differently from other review tools by focusing on both successful and failed experiments. It learns from published null results, failed studies, and other evidence that doesn’t support claims. The system assigns scores to each claim in a paper based on originality and validity.

The goal of the top-1% list was to highlight research quality regardless of author or journal prestige. QED Science found 12.9% of papers they reviewed were “hidden gems” missed by traditional publishing systems. Experts agreed with their rankings in 75% of cases.

When asked if the tool could replace peer reviewers, QED Science said it aims to help researchers improve their work before publication, not replace peer review or be used by publishers.

The tool’s role is to assist researchers in enhancing their manuscripts privately and securely, focusing on the science itself rather than external factors like journal reputation.