GPTZero has raised concerns about a KPMG report on agentic AI, saying the document contained citation problems that looked like the kind of errors often associated with generative AI output.
The claim has not been independently verified here, so the point is best read as an allegation from GPTZero rather than a settled finding. Still, the case is a useful warning for anyone relying on consultant reports, vendor research, or market studies to make decisions about AI tools and strategy.
According to GPTZero, only five of the report’s 45 citations accurately pointed to real sources. The company said other references appeared to be fabricated, distorted, wrongly attributed, or attached to titles that did not match the underlying material.
What GPTZero Is Calling “Vibe Citing”
GPTZero used the phrase “vibe citing” to describe references that look credible at first glance but do not reliably connect to real or accurately described sources. In its view, that can include made-up citations, garbled source details, paraphrased titles, or cases where topics and authors appear to be confused.
The company argued that the pattern matters because a polished report can travel far beyond its original audience. A document from a major consulting firm may be quoted in news coverage, blog posts, procurement discussions, investor notes, presentations, and internal business plans. AI hallucinations If flawed citations are repeated in those settings, readers may treat weak evidence as stronger than it is.
GPTZero also suggested that this kind of material could be absorbed into large language model outputs if it is cited, scraped, or summarized elsewhere. That concern has not been independently confirmed here, but it reflects a broader worry: once a questionable claim is published in a professional-looking format, it can become harder to trace and correct.
Why This Matters For AI Buyers
For companies evaluating agentic AI, the practical lesson is not limited to one report. Research from large consultancies can be useful, but it should not replace basic source checking, especially when the report is being used to justify budget, vendor selection, compliance planning, or operational change.
Before leaning on an AI report, buyers should check whether the most important claims are tied to sources that actually exist, whether those sources say what the report says they say, and whether case studies are specific enough to verify. A citation list is not the same thing as evidence if the references cannot be followed cleanly.
This is also a reminder that AI-generated writing can sound confident while still failing at the parts of research that matter most: accurate attribution, careful sourcing, and clear separation between evidence and interpretation.
The Bottom Line
GPTZero’s allegations about the KPMG AI report remain framed here as claims, not independently verified facts. Even so, the episode points to a real buyer-side risk. As AI research becomes more influential, readers need to judge reports not only by the brand on the cover, but by whether the underlying evidence holds up.