PwC Caught Using AI-Generated Reports With Fake Citations and Fabricated Claims

PwC Middle East Publishes Reports Riddled With AI Hallucinations

PwC has become the latest Big Four firm caught publishing thought leadership reports marred by AI hallucinations, fake footnotes, and fabricated citations1

1

2

2

. The discovery was made by GPTZero researchers and verified by the Financial Times, revealing that PwC Middle East produced four reports between 2024 and 2026 containing extensive AI-generated inaccuracies1

1

. These consulting reports were designed to attract clients in the Middle East, covering topics including agentic AI bots, electric vehicles, and government public services3

3

.

The AI misuse in professional services is particularly embarrassing for PwC, which markets itself as an adviser on responsible AI implementation and policies to avoid exactly these types of errors1

1

2

2

. One report claimed that 70% of Middle Eastern chief executives believed generative AI would significantly affect their business, but the accompanying footnote linked to a media report making no mention of the survey1

1

2

2

.

Blatant Evidence of AI-Generated Content

The most glaring evidence came from a cybersecurity report whose footnote URL included the tracking tag "utm_source=chatgpt.com," leaving little doubt about its origin2

2

3

3

. Another egregious example involved a report citing a teenage blogger with 280 followers on Medium as the sole source for what PwC called a "real-world success story" about a JPMorgan agentic AI initiative1

1

4

4

. The initiative, which automated commercial loan agreement reviews and saved hundreds of thousands of hours of human work, actually occurred in 2017—five years before ChatGPT launched1

1

2

2

.

GPTZero researcher Paul Esau identified telltale patterns of AI-generated reports throughout the documents1

1

. One report cited the claim that human error causes 90% of traffic accidents three times across two pages, using three different sources—something "no human is going to" do, Esau explained1

1

4

4

. The chaotic signposting of source material is symptomatic of AI-generated research, he noted2

2

.

Fabricated Citations and Nonexistent Sources

Many fabricated citations in the AI-generated reports linked to web pages that either didn't contain the purported evidence or didn't exist at all1

1

. A 2025 report included an entire section about a framework called "Citizen Pulse," allegedly deployed by governments worldwide4

4

5

5

. GPTZero discovered the framework didn't actually exist and was entirely hallucinated by AI, including supporting citations4

4

5

5

.

An academic paper on air quality in Riyadh appears to have been completely fabricated, with no trace of the study in the referenced journal or by the attributed authors1

1

. These fake footnotes and AI slop undermine the credibility of firms charging premium rates for judgment and rigor3

3

.

Fourth Big Four Firm Caught Publishing AI Slop

PwC completes an unwanted set among the Big Four consulting firms3

3

. GPTZero's earlier investigations led KPMG to retract an October report containing false claims about AI use at UBS, the UK's NHS, and Transport for London1

1

3

3

. EY retracted a study last month over fake footnotes, while Deloitte has twice included AI fabrications in government reports, including a $1.6 million health plan in Canada citing papers that don't exist3

3

5

5

.

The pattern repeats each time: Big Four firms pump out hundreds of AI thought-leadership pieces to win clients while pushing staff to use AI to work faster, resulting in AI slop with blue-chip logos3

3

. The timing proved particularly awkward for PwC, as the same week these revelations emerged, PwC's US chief executive Paul Griggs warned companies about AI misuse, stating that "all AI is going to do for you is tell you how bad your messy process really is"3

3

.

PwC's Response and Quality Control Failures

PwC Middle East responded to the findings with a measured statement, saying it "takes the accuracy of our published research seriously and is updating a limited number of supporting citations" in the identified reports1

1

2

2

. The firm added that it has quality control processes for research and content development that all staff should follow, though it failed to address how the errors made it into published reports1

1

5

5

.

The discovery raises uncomfortable questions about the quality of professional services when prestigious consulting firms provide clients with unverified AI-generated content4

4

. These firms charge premium rates while marketing themselves as authorities on responsible AI adoption, yet their own quality control processes failed to catch obvious AI hallucinations before publication2

2

3

3

. The gap between what consultants tell clients and their own practices reveals how hype can outrun safe use, and how AI-generated output that looks right often goes unchecked3

3

.