
AI Text Detectors Fail in the Wild. Now We Know Why.
Most AI text detectors look good in testing. They’re trained and evaluated on specific models, specific prompt styles, specific domains — and they perform well within that distribution. Then they
Booth 21-25 | AI Data Management Zone | Tokyo Big Sight

Most AI text detectors look good in testing. They’re trained and evaluated on specific models, specific prompt styles, specific domains — and they perform well within that distribution. Then they

A 34-author survey led by Yu Xia at UC San Diego — with collaborators across Adobe Research and nine other labs — was accepted to ACL 2025, and it’s worth your time if you build data pipelines. The title carries the whole argument: “From Selection to Generation.” For most of
