KPMG AI report found riddled with fake citations

TL;DR:

• A GPTZero investigation found that only five of the 45 citations in KPMG’s October 2025 agentic-AI report pointed to real, intact sources.
• Around half of the factual claims were false or misattributed, and the report contradicted KPMG’s own CEO survey figures.
• It is the fourth professional-services firm in recent months — after Deloitte, EY and two law firms — to be caught publishing AI-fabricated content.

A report by KPMG extolling the promise of agentic AI has turned out to be a cautionary tale about it. AI-detection firm GPTZero examined the Big Four firm’s October 2025 study, Total Experience: Redefining Excellence in the Age of Agentic AI, and found that 40 of its 45 citation titles were fabricated. KPMG has since pulled the report from its homepage.

When the source itself is hallucinated

GPTZero’s analysis described the firm “vibe citing” — generative tools inventing references, blending real sources, or mangling titles beyond recognition. One citation offered a 2019 railway press release as evidence of AI-agent use, despite the term entering common usage only in 2024. The model also repeatedly confused article subjects with authors. More awkwardly, the report cited “KPMG research” claiming 55% of chief executives rank AI as their top investment priority, while KPMG’s own 2025 CEO Outlook, published the same month, put the figure at 71%.

The flawed statistics have already spread, picked up by trade publications and now surfacing in answers from ChatGPT and Gemini — a small illustration of how a single unverified document can pollute the wider information supply.

For UK businesses weighing which advisers to trust on AI, the pattern is the worrying part. Deloitte refunded part of an Australian government contract over AI errors; law firm Pinsent Masons drew High Court criticism after a lawyer filed AI-generated letters with false legal content; and EY withdrew a loyalty-rewards study containing invented footnotes. The firms selling AI assurance keep tripping over the same failure.

Looking forward

The common thread is absent human review, not the technology itself. As the Chartered Management Institute’s latest research on the UK management skills gap underlines, governance and oversight are lagging adoption. Until firms can show their own published work survives a fact-check, “responsible AI” advice from the Big Four will be read with a sceptical eye.