An investigation by GPTZero has revealed that PwC Middle East’s AI-assisted reports contained fabricated references, raising fresh concerns about the reliability of AI-generated content in business and government advisory work.
Readers attempting to verify the sources found that many citations were untraceable, with broken links and references to academic studies that never existed. In one case, a URL contained the telltale utm_source=chatgpt.com parameter, indicating the text was generated by ChatGPT.
PwC Middle East has acknowledged the issue and stated it is reviewing each report and conducting thorough fact-checking. The firm is not alone: GPTZero previously flagged similar AI-generated reports from KPMG and EY that also contained hallucinations.
This incident underscores a critical weakness of generative AI: systems can produce fluent, convincing text but often invent facts or references when not carefully verified. Experts stress that AI should augment human judgment, not replace it, and that robust fact-checking remains essential.