The decision by ACL exposes a growing integrity problem in AI research publishing, where the use of large language models in literature review is producing fabricated references that survive peer review.