TL;DR

PwC published four “thought leadership” reports containing fake footnotes, misattributed claims and unverifiable sources, according to an investigation by researchers at GPTZero verified by the Financial Times. The reports — covering agentic AI for corporates, public service improvement for governments, and electric and autonomous vehicle forecasts — were produced by PwC Middle East over the past two years to generate consulting work.

The failures are specific. Many citations link to pages that do not contain the evidence claimed, and some to pages that do not exist at all. An academic paper on air quality in Riyadh appears to have been fabricated outright: there is no trace of the study in the journal cited, nor from the authors it is attributed to.

Other footnotes are merely odd. A teenage blogger with 280 followers on Medium is given as the source for a JPMorgan initiative PwC describes as a real-world success story of agentic AI. The initiative in question — automating the review of commercial loan agreements — was first reported in 2017, five years before ChatGPT launched generative AI into mainstream use.

Paul Esau, a researcher at GPTZero, pointed to a structural tell rather than a factual one. A claim that human error causes 90% of traffic accidents appears three times in one report: first with a footnote, then without, then with two footnotes citing different sources. “This claim isn’t fake, but no human is going to cite the same fact three times in two pages using three different sources,” he said. One report cites a PwC survey of Middle Eastern chief executives but footnotes a media report that never mentions it. Another footnote carries the URL parameter “utm_source=chatgpt.com”.

PwC Middle East said it takes the accuracy of its published research seriously and is updating a limited number of supporting citations, without explaining how the errors got in.

GPTZero’s earlier work led EY and KPMG to retract reports for the same reason.

Looking Forward

The awkwardness for UK firms is not the errors themselves but where they sit. These are documents marketing PwC’s expertise in helping clients adopt AI responsibly, including how to implement policies that prevent exactly this. Three of the Big Four have now been caught the same way, which suggests the gap is between AI usage policies and any mechanism for verifying citations before publication — a check that costs little and evidently is not happening.