PwC has come under scrutiny over the use of artificial intelligence (AI) in its research publications after an investigation identified fabricated citations, broken links and references to studies that appear to be nonexistent.
According to the Financial Times, researchers from AI detection platform GPTZero reviewed four thought leadership reports published by PwC Middle East over the past two years. Their analysis identified several instances of AI-generated “hallucinations” across reports covering topics such as corporate strategies for “agentic” AI, government services and electric vehicle market projections in the Middle East.
The researchers found problems with multiple citations. One report referenced an alleged study on air quality in Riyadh that could not be located in the journal or under the authors cited. Elsewhere, footnotes directed readers to broken webpages or articles that did not contain evidence supporting the claims attributed to them.
Another report cited a teenage Medium blogger with roughly 280 followers as evidence for a real-world AI success story involving JPMorgan. The Financial Times reported that the initiative in question actually took place in 2017, several years before ChatGPT was introduced.
Researchers also discovered that an identical claim concerning traffic accidents appeared three times across two pages, with a different source attached to each occurrence. In addition, one cybersecurity report cited a URL containing the tracking parameter “utm_source=chatgpt.com”, indicating that the source may have been discovered through ChatGPT.
In response, PwC Middle East told the Financial Times that it "takes the accuracy of our published research seriously" and said it was updating a limited number of supporting citations in the affected reports.
PwC Joins Other Big Four Firms Facing AI-Related Errors
PwC is the latest major consulting firm to face questions over the accuracy of AI-assisted research.
Earlier this year, KPMG pulled its global report, "Redefining Excellence in the Age of Agentic AI", from publication after companies and organisations referenced in the document challenged several of the case studies and achievements attributed to them. The Financial Times reported that the report included false claims and fabricated case studies generated with AI.
EY Canada also removed a report examining loyalty rewards programmes after researchers identified AI hallucinations, incorrect citations and references to studies that could not be verified. The firm subsequently said it was reviewing the circumstances surrounding the report's publication.
Deloitte has faced similar scrutiny after producing a healthcare report for the Canadian government that reportedly contained several factual inaccuracies believed to have resulted from AI tools, according to media reports cited by the Financial Times.
AI Oversight Comes Under Increasing Scrutiny
The incidents involving PwC and its Big Four peers are intensifying questions about how consulting firms use generative AI while advising businesses and governments on AI adoption, governance and risk management.
The repeated errors highlight concerns around editorial controls, source verification and human oversight before AI-assisted content is released publicly. These safeguards become particularly important when research is used to influence corporate strategies, investment decisions or public policy.
Although consulting firms increasingly position AI as a tool for improving productivity and efficiency, the experiences of PwC, KPMG, EY and Deloitte demonstrate the potential risks of publishing AI-assisted research without thorough human fact-checking.
The situation also presents an uncomfortable contradiction for the consulting sector: firms advising governments and global businesses on responsible AI adoption are simultaneously being challenged over their ability to ensure that AI-assisted content in their own publications is accurate, properly sourced and independently verified.