Do failed replications make their way into public knowledge? | Youyou Wu

Youyou Wu asks whether failed replications reach public knowledge. Using FORRT's replication database, the study selects findings whose replications all failed, with large samples and peer-reviewed publication, and checks whether Wikipedia, textbooks and course syllabi have updated their mentions: 36%, 27% and 35% have, more than an audience poll expected. ChatGPT mentions the replication evidence in 80% of answers to lay prompts and 67% to expert prompts, Google's AI Overview in 42%; without web search ChatGPT drops to 40%. Rates are lower for findings tested in large replication projects and higher for findings listed on FORRT's Reversals & Replications page. Wu closes with advice for curators, researchers and readers, and notes that textbooks tend to drop a section rather than revise its narrative. Speaker: Youyou Wu (University College London) Authors: Youyou Wu, Yang Yang and Brian Uzzi Talk title: Do failed replications make their way into public knowledge? Session: Session 5 · Reproducing the record Recorded at AI in Metascience 2026, an online conference organised by FORRT (Framework for Open and Reproducible Research Training), 28–30 September 2026. Day 2, Tuesday 29 September: Replications and reproductions. Programme and abstracts: https://forrt.org/ai-metascience/ FORRT: https://forrt.org