EY Retracts Loyalty Study After GPTZero Exposes AI-Fabricated Citations

EY, one of the world's Big Four accounting and consulting firms, has been forced to retract a cybersecurity study on loyalty programs after AI detection company GPTZero exposed widespread fabricated citations and hallucinated data throughout the report. The retraction raises serious questions about quality control as major enterprises increasingly rely on artificial intelligence tools to produce professional research.

What GPTZero Found in EY’s Report

GPTZero, a company that specializes in detecting content generated by large language models (LLMs — the technology that powers tools like ChatGPT), investigated EY Canada’s 44-page report on cybersecurity threats in loyalty reward systems. The results were damning. Roughly 60 percent of the references in the report were hallucinated, meaning the AI tool used to help write it invented sources that simply do not exist. One fabricated citation pointed to a McKinsey report that was never published anywhere. In separate sections, the document cited contradictory figures, estimating both the total loyalty program market size and the volume of unredeemed loyalty points at exactly 200 billion dollars. The report was credited to three EY employees, including two partners and one senior manager.

Why This Matters for Enterprise AI Adoption

The incident highlights a growing concern as businesses rush to integrate generative AI tools into professional workflows. When a firm with EY’s global reputation publishes research riddled with invented data, it undermines the credibility that consulting firms depend on to justify premium fees. EY quickly removed the report from its website and stated it was reviewing how the publication occurred, emphasizing the study was not connected to any client work.

This is far from an isolated case. Multiple professional services firms have faced similar embarrassments as employees use AI writing assistants to draft reports without adequate fact-checking. The core problem is that LLMs generate text that reads confidently and professionally, even when the underlying claims are entirely fabricated. Without rigorous human verification at every step, these hallucinations can slip through existing review processes that were designed for human-written content.

The EY retraction serves as a cautionary tale for any organization deploying AI in research and publishing. Quality control must evolve to match the unique risks of AI-generated work, where the biggest threat is not poor writing but convincingly false information.