Ernst & Young (EY) withdrew a study on loyalty rewards programs after researchers discovered AI-generated hallucinations and fabricated footnotes [1].
The incident highlights the risks of integrating generative AI into professional consulting and the potential for high-profile firms to publish unreliable data. It underscores a growing tension between the speed of AI adoption and the necessity of rigorous human oversight in corporate research.
Researchers at GPTZero identified the apparent hallucinations and fake citations within the report [1]. The findings prompted EY to remove the document from its Canada website [4]. The firm said it is reviewing the circumstances that led to the publication of the flawed study [4].
According to reports, the retraction occurred on May 17, 2026 [2]. The firm said it is examining its internal publication processes to prevent similar occurrences in the future [4].
AI hallucinations occur when a large language model generates confident responses that are factually incorrect or entirely invented. In this instance, the AI did not simply misinterpret data but created footnotes that did not exist in any real-world source [1].
EY has not provided a detailed explanation of which specific AI tools were used to generate the report, but the company said it is reviewing how the content was vetted before it reached the public domain [4].
“EY withdrew a study on loyalty rewards programs after researchers discovered AI‑generated hallucinations.”
This retraction serves as a cautionary tale for the professional services industry, where accuracy is the primary product. As firms like EY integrate generative AI to scale research, the discovery of fabricated citations suggests that traditional peer-review and fact-checking layers may be insufficient to catch sophisticated AI errors. This may lead to more stringent auditing requirements for AI-assisted corporate reports.
