Top Consulting Group PwC Caught Using AI on “Thought Leadership” Report About AI, Resulting in Corporate Document Filled With Bizarre Hallucinations

PwC, a prominent consulting firm among the "Big Four," has been exposed for publishing "thought leadership" reports on AI that are riddled with inaccurate and easily identifiable fabrications. According to the Financial Times, four such reports, published between 2024 and 2026, were intended to generate consulting business in the Middle East.

However, AI detector company GPTZero discovered the reports to be riddled with "hallucinated citations, fabricated claims, and confusing formatting choices." Contents of these reports included "claims that are unrelated to or contradict the cited sources." For example, a 2025 report contained an entire segment on a concept called "Citizen Pulse," which the software revealed was fabricated and did not exist in reality.

The findings shed light on the growing prevalence of AI hallucinations, which range from lawyers being reprimanded for including fictitious legal citations to entire academic disciplines being inundated with dubious AI-generated research. The consulting industry, too, has fallen victim to this issue, as seen when KPMG was caught inserting numerous AI-generated inaccuracies in a report on integrating AI agents in investment advisory, risk management, and compliance monitoring.

What makes this recent case particularly damning is that several of PwC's poorly researched reports were directly about "agentic" AI bots, such as a guide for governments on enhancing public services for autonomous vehicles. The extent of the problem is astonishing. One PwC report cited a teenage blogger from Medium with only 280 followers as a source for a JPMorgan initiative on agentic AI.

GPTZero analyst Paul Esau found another PwC report that cited the same correct claim about human error causing 90 percent of traffic accidents three times in just two pages, using three different sources—an occurrence that "no human" would ever do. In response, PwC downplayed the findings and did not accept responsibility. In a statement to the newspaper, PwC Middle East claimed it "takes the accuracy of our published research seriously and is updating a limited number of supporting citations" in the reports.

"Consistent with our approach to responsible AI, we have quality control processes for research and content development we expect all our people to adhere to."