A major KPMG report on agentic AI has been found to be filled with AI-generated errors, false citations, and misleading case studies, according to an investigation by GPTZero. The findings highlight a growing problem of AI hallucinations infiltrating authoritative documents.
The Evidence of Hallucinations
GPTZero investigators examined the KPMG report and discovered that only five of the 45 citations accurately pointed to real sources. The remaining citations were either totally false or significantly distorted — a pattern the company calls 'vibe citing', where generative AI creates plausible-looking but fake references.
The team noted that the report included odd mixes of real references, such as wrong attributions or paraphrased titles. "A human would not consistently paraphrase titles, mistake topics for authors or repeat information across multiple components," they wrote.
Consequences of Vibe Citing
Although the researchers considered arguments for and against vibe citing, they ultimately concluded it should still be considered hallucination and that "vibes have consequences." In KPMG's case, the firm's influence means its findings are likely to be cited globally across news reports, blog posts, and other conversations, driving the dissemination of potential misinformation. GPTZero also warned that the report was being cited in LLMs, spreading the false information even further.
| Citation Status | Count |
|---|---|
| Accurate real sources | 5 |
| Totally false or distorted | 40 |
A Broader Trend
This is not an isolated incident. A similar 2025 report revealed that a study from the US Presidential Commission to Make America Healthy Again (MAHA) also included "garbled or fabricated" footnotes, reports TechRadar.
Implications for Enterprise Decision-Makers
GPTZero contends that vibe citations are a clear and present danger to researchers, academics, consultants, students, and anybody who happens to search the internet for information. For enterprise technology buyers, this case underscores the critical need to verify sources in AI-related reports before basing procurement or strategy decisions on them. The KPMG report, despite its flawed citations, could still be referenced by other AI systems and human analysts, making the spread of inaccuracies difficult to contain.
As generative AI becomes more embedded in research and consulting, the risk of hallucinated content multiplying through authoritative channels grows. The message from GPTZero is clear: AI-generated outputs must be treated with skepticism, and rigorous validation is non-negotiable.
Sources:
