GPTZero Investigation Exposes Widespread Hallucinated Citations in KPMG's Agentic AI Report
Key Takeaways
- ▸KPMG's 'Total Experience' report contains 40 of 45 fabricated or paraphrased citation titles, with only 5 citations accurately pointing to real sources
- ▸Nearly half of the claims supported by citations appear to be fake or misattributed, suggesting LLM hallucinations during the research process
- ▸GPTZero introduces 'vibe citations' to describe hallucinations on a severity spectrum—from fully fabricated references to paraphrased or fused combinations of real sources
Summary
GPTZero investigators used their Hallucination Check tool to analyze KPMG's October 2025 report, 'Total Experience: Redefining Excellence in the Age of Agentic AI,' which examines consumer experience around the world. The analysis uncovered a troubling pattern: of 45 citations in the report, only 5 accurately point to real sources, while 28 contain paraphrased titles or fake components for real sources, and 12 are too vague to verify. Additionally, approximately 40 of the 45 citation titles appear to be fabricated or heavily paraphrased.
The investigation introduces the concept of 'vibe citations'—hallucinations generated by AI research tools on a spectrum of severity, ranging from completely fabricated references (fake authors, titles, and publishers) to fusions of multiple real sources or paraphrased versions of legitimate citations. Around half of the claims evidenced by the citations appear to be fake or misattributed, likely resulting from an AI tool over-complying with a request to find examples of 'agentic AI' in practice.
This case represents the second in a series of GPTZero investigations into hallucinated citations in high-profile documents, following previous discoveries in government reports, academic papers from prestigious ML/AI conferences (ICLR, NeurIPS), and reports from other Big Four consulting firms including Deloitte and Ernst & Young. The findings raise critical questions about the integrity of AI-generated research and the adequacy of current citation verification practices in professional consulting.
- Systemic issues with AI-generated research in major consulting firms and academic institutions pose risks to research integrity and informed decision-making
Editorial Opinion
This investigation exposes a critical vulnerability in professional consulting: the uncritical integration of AI research tools without robust fact-checking mechanisms. As enterprises increasingly rely on AI-generated reports for strategic decision-making, the prevalence of 'vibe citations' in KPMG's analysis raises urgent questions about AI governance standards. Organizations must implement rigorous citation verification processes and mandate transparency about AI tool usage in research workflows—otherwise, hallucinations risk becoming normalized in high-stakes knowledge work.
