PwC fabricated sources allegations hit Middle East AI reports

GPTZero says it found PwC fabricated sources and unsupported claims in four PwC Middle East reports published between 2024 and 2026, putting another Big Four firm under scrutiny for AI-era research controls. No transaction value or customer revenue is involved, but the finding matters because PwC used at least one of the reports to market a product to government clients.

The AI-detection company said the most severe case was a PwC Middle East report called "Transforming Governance," which it rated as 84% likely to have been entirely generated by AI. According to GPTZero, that report promoted a product called Citizen Pulse and said governments in Denmark, Saudi Arabia, the United States and Australia rely on it. GPTZero said it found no evidence backing those adoption claims.

Did PwC publish AI-generated reports with fabricated sources?

GPTZero's finding is an allegation based on its analysis of the reports and citations. The company said the PwC Middle East reports with the highest AI scores also had the largest number of false or unsupported references.

GPTZero described the pattern as "Vibe Citing," meaning citations that appear to support a claim but are loosely attached, incorrectly described or unusable because titles, authors or URLs are wrong. In the cases it reviewed, GPTZero said some references did not support the claims they were paired with.

PwC Middle East told the Financial Times that it takes accuracy seriously and is updating a limited number of supporting citations. The firm did not explain to the Financial Times how the citation problems entered the reports.

Why the citation issue matters for consulting firms

Consulting firms sell trust as much as advice. For AI products and public-sector transformation work, reports often function as sales material, proof of expertise and market education at the same time. If citations are fabricated or claims about government use cannot be verified, the problem is larger than formatting.

GPTZero has previously reported fabricated-source issues involving KPMG, Deloitte and Ernst & Young. The PwC allegations therefore land as part of a broader pattern, according to GPTZero, rather than as an isolated document cleanup problem.

The open question is whether the errors came from generative AI, weak editorial review, human sloppiness or some mix of those factors. GPTZero's analysis links higher AI-likelihood scores with more citation problems, but PwC Middle East has not publicly described the production process for the reports or said whether generative AI was used.

For enterprise buyers, the practical issue is less philosophical: if a firm is selling AI-enabled governance or public-sector modernization, its own publishing workflow becomes evidence of how well it handles accuracy, review and accountability. PwC Middle East's response so far addresses some citations, but not the process that produced them.

This story draws on original reporting from The Decoder.