PwC and other Big Four firms exposed for AI-generated errors in research reports
An audit by GPTZero found that PwC Middle East and other Big Four consultancies published reports containing fabricated citations and AI-hallucinated content, including a footnote that directly referenced ChatGPT.
An investigation by GPTZero, corroborated, uncovered AI-driven mistakes in four PwC Middle East thought-leadership reports issued in the last two years. The audit revealed fabricated footnotes, nonexistent journal articles, broken URLs and even a citation that listed ChatGPT in the source address, exposing the use of generative tools to draft the research. Among the errors were a phantom Riyadh air-quality study, a reference to a teenage Medium blogger for a JPMorgan AI case that pre-dated ChatGPT, and duplicated traffic-accident claims each backed by different sources.
PwC Middle East responded that it takes research accuracy seriously and will update the flawed citations. The episode follows earlier scandals where GPTZero found similar hallucinations in reports from EY Canada, KPMG’s global “Agentic AI” paper and a Deloitte healthcare assessment for the Canadian government, prompting retractions and reviews across the Big Four.
Why it matters
The story shows how reliance on generative AI can undermine credibility of major consulting firms and mislead clients.
In this story