Abstract
Retractions reveal failures the record otherwise hides. In 2012, a landmark PNAS study showed that misconduct, not error, accounted for most retracted papers. This Brief Report updates that signal for the generative AI era. Using Retraction Watch records through May 2026, it shows that retractions tagged as computer-aided or computer-generated content did not disappear after the 2023 Hindawi cleanup. In 2025, with the Hindawi contribution at zero, the AI-content share of all retractions remained at one-quarter (23.6%), from other publishers. The finding matters because it separates a one-publisher event from a wider publishing-system condition that researchers, editors, and funders should recognize: computer-generated content is now a durable documented cause of retraction, not merely an artifact of a single publisher's record.