EMNLP Partners With GPTZero to Reject Papers With Hallucinated References
EMNLP will use GPTZero to screen the references in each paper submitted, which is a huge step forward in the fight against hallucinated academic citations.
EMNLP, also known as the Conference on Empirical Methods in Natural Language Processing, is a leading conference in the area of natural language processing and artificial intelligence. It’s well-known in the industry, and researchers from universities and major AI companies, including OpenAI and Anthropic, submit papers.
We’re proud to announce a new partnership to help them detect references that do not exist in academic databases. This means that our hallucination detector will be applied to every entry in the bibliography of every submitted paper.
Flagged entries will be further examined and verified by the senior program committee (AC/SAC/PC chairs) of the submission. Papers with any unverifiable reference may be desk rejected, and all authors involved in the submission may be ineligible to submit their paper(s) to EMNLP 2026 and EMNLP 2027.
This means a single hallucinated reference could result in a paper being rejected before it reaches the full peer-review process, and it also means that responsibility sits with every author named on the paper.
The danger of hallucinated citations is becoming increasingly recognised, as after a false citation appears in an accepted paper, future researchers may repeat it because they assume it has already been checked. We’re very familiar with the issue, and in December 2025, we used our Hallucination Check tool to scan a sample of 300 ICLR 2026 papers submitted to OpenReview.
Our tool flagged 90 papers as containing at least one citation that appeared not to exist online. Following human verification, we determined that 50 papers included at least one actual hallucinated citation. Then, in January 2026, as reported in Fortune, we found at least 100 hallucinations in accepted NeurIPS 2025 papers, NeurIPS being one of the world’s most prestigious machine learning conferences.
We’re proud to partner with EMNLP and to support its work to keep unverifiable references out of the academic record. This continues our mission to preserve what’s human and to be building an authenticity layer for the internet.
