EY Canada takes down study after apparent AI hallucinations

EY Canada has removed a study on loyalty rewards programs from its website after GPTZero, a leading AI detector, found evidence of AI hallucinations and false footnotes.

GPTZero’s investigations branch on May 14 published a report that found 16 of 27 references in the EY study were hallucinated and that 72% of the study was AI.

GPTZero – which offers an AI detector solution that is used by hundreds of organizations in education, hiring, publishing, and legal – found that more than a half-dozen footnotes led to dead webpages or did not contain the cited information. The erroneous data in the EY study included references to a fabricated McKinsey report that claimed US$200 million in unredeemed loyalty rewards globally.

Following the publication of the GPTZero report, EY Canada removed the study “Points of Attack: Uncovering Cyber Threats and Fraud in Loyalty Systems” from its website. The Big Four accountancy said it was reviewing why the study was published.

AI-generated research with “vibe citations” can be problematic, GPTZero says, especially if it is published by well-known consultancies with high-traffic sites. Studies with made-up figures can “poison the well” for future researchers.

AI-hallucinated research is also a threat to the credibility and business models of consultancies that have gone full-tilt on AI integration. If a client receives AI-regurgitated recommendations from a consulting firm, what exactly is the point of hiring EY instead of buying a Claude license?

Fellow Big Four firm Deloitte was lambasted last year for providing the Newfoundland government with a $1.6-million healthcare report that contained AI-generated errors and fake academic citations.