The Association for Computational Linguistics (ACL) has desk-rejected more than 100 papers accepted to its 2026 annual conference after finding citations to non-existent literature in their camera-ready versions.
ACL is the primary international professional organization for researchers working on natural language processing and computational linguistics, making its annual conference one of the most cited venues in AI research. A hallucinated citation entering ACL proceedings does not stay there, it gets cited by future papers, indexed by databases, and treated as real, which is how fabricated knowledge propagates through the scientific record.
The rejections were announced by the ACL 2026 program committee chairs in a statement published on the conference website. The papers had already cleared peer review and were being prepared for publication when the issue was found. An automated system flagged potential citation problems, and human experts, described as program chairs and senior area chairs, reviewed each flagged citation to confirm it did not exist.
The program chairs said the finding applied to papers already accepted and no longer subject to anonymity requirements. A follow-up note on the statement said affected authors must go through the normal resubmission process.
The statement did not specify how many of the rejected papers used large language models to generate citations. It said the cause was immaterial. “An author who fabricates a citation commits a serious breach of ethics, and using an automated system as a proxy to generate such citations is equally unacceptable,” the chairs wrote.
ACL and the ACL Rolling Review system had previously required all citations to be verified and real. The chairs acknowledged that some authors may have already made travel arrangements for the conference.
ACL 2026 is scheduled to be held this year. The 64th annual meeting is organized by program chairs David Jurgens, Maria Liakata, Viviane Moreira, and Jiajun Zhang, their statement said.
Also Read: Study Reveals LLMs generated 150,000 Fake Citations in Research Papers
