PwC is the third Big Four firm caught shipping AI-hallucinated reports
GPTZero found fabricated citations and a framework 'used' by four governments that can't be verified. KPMG and EY were caught first — here's what to check.
Copy markdown
KPMG, EY, and now PwC
GPTZero says four PwC Middle East 'thought leadership' reports from 2024-2026 are AI-generated or mixed and riddled with fabricated citations. It's the same detector that forced KPMG to pull a report in June and pushed EY to retract others — the failure has now hit three of the Big Four.
The fabrications, specifically
One report pitches a 'Citizen Pulse' framework it claims governments in Denmark, Saudi Arabia, the US and Australia already run — GPTZero found no public evidence it exists. It also cites a nonexistent MIT Technology Review article and a WEF report whose title is invented, and one chart conflates 13M EV sales with 3.58M charging ports.
The scores, in numbers
The four reports rated 100%, 82%, 60% and 50% AI-generated-or-mixed, across 17, 43, 19 and 32 citations. GPTZero's takeaway: the more AI-written the document, the more of its sources turn out to be fabricated.
The tells you can check yourself
The flags are reproducible in minutes: source URLs carrying utm_source=chatgpt.com, HTML metadata titles that don't match the visible ones, and citations that don't actually support the claim attached to them. Run those three checks on anything an LLM drafted before it ships.
Why this is your problem too
If you generate client-facing docs, marketing, or report copy with a model, this is your failure mode: invented customers and fake citations survive casual review and surface in public. A citation-verification pass — GPTZero's Hallucination Check or your own source-matching step — is now table stakes, not polish.