KPMG pulls AI report after investigation finds it was riddled with AI hallucinations and fake citations
A widely circulated KPMG report championing the use of agentic AI was withdrawn from some websites after GPTZero investigators discovered that half of its claims were fake or misattributed and only five of its 45 citations pointed to real sources.
What KPMG published
In October last year, KPMG published a report titled Total Experience: Redefining Excellence in the Age of Agentic AI, which detailed how companies were using artificial intelligence to serve customers. The consultancy, one of the Big Four professional services firms along with Deloitte, PricewaterhouseCoopers and Ernst & Young, positioned the paper as a trusted industry survey. According to the Financial Times, the report was later withdrawn from some websites after KPMG was notified of the problems.
What the investigation found
Investigators for GPTZero, a maker of AI content detection tools, examined the report and discovered that inaccuracies and fake footnotes were pervasive. Of the 45 citations in the document, only five accurately pointed to real sources. Twenty-eight citations paraphrased titles or added fabricated components to genuine references, while twelve were phrased too vaguely to determine whether they existed at all. GPTZero coined the term “vibe citing” for the phenomenon of AI models generating plausible-looking but fake references. The team also concluded that roughly half of the substantive claims in the report were either false or misattributed, likely the result of an AI research tool over-complying with a prompt to find examples of agentic AI.
- Accurate
- 5 citations
- Paraphrased or partly fabricated
- 28 citations
- Too vague to verify
- 12 citations
Three disputed case studies
Several organisations mentioned in the report have publicly corrected the record. KPMG claimed that Emirates launched a mobile chatbot called Sara that could talk to passengers and alter their flights. In reality, Sara was a mobile assistant introduced in 2023, not an AI-powered chatbot, and it never had the capability to change bookings. KPMG also wrote that UBS had integrated agentic AI across its investment advisory, risk management and compliance monitoring functions. A UBS spokesperson told the Financial Times that the information was “factually incorrect.” Swiss Federal Railways (SBB) was said to employ AI agents that help passengers plan, book and optimise trips based on preferences, real-time conditions and carbon impact; an SBB spokesperson said the claim was “not accurate.” The Irish Times added that the report also contained a false case study concerning the UK’s National Health Service.
Wider consequences
GPTZero chief executive Edward Tian warned that error-riddled papers published by the Big Four could “poison the well of information.” Because reports from firms like KPMG are treated as highly trusted sources, they are routinely cited in other research papers, news articles and blog posts. GPTZero’s latest investigation also found that some of the discredited content was already being ingested by large language models, risking further spread. The episode follows a similar 2025 revelation in which a study from the US Presidential Commission to Make America Healthy Again contained garbled or fabricated footnotes.
Vibes have consequences.
The Irish Times noted that the incident points to a structural problem: AI allows consultancies to produce reports quickly, but when nobody checks the facts, the resulting documents can amplify misinformation across the knowledge ecosystem.
