KPMG Pulls AI Report After 40 of 45 Citations Fail Check

KPMG International pulled its October 2025 report "Total Experience: Redefining Excellence in the Age of Agentic AI" from parts of its website this week after a forensic review by detection company GPTZero found that only five of the report's 45 citations correctly pointed to the source they named. The Big Four firm confirmed the removal as it reviews how the document was produced, according to TechCrunch's June 13 report.

The report leaned on case studies of agentic AI deployments at UBS, the UK National Health Service, Swiss Federal Railways and Transport for London. All four organizations told KPMG that the descriptions of their AI use were either wrong or misleading, and UBS publicly pushed back on the claims about the bank.

GPTZero's audit, summarized by The Register on June 12, classified 28 of the 45 citations as paraphrased or fabricated and 12 as too vague to verify. The reviewers described what they called "vibe citing," in which a generative tool stitches together fragments and invents convincing but false references. The report also stated that 55 percent of CEOs ranked AI as a top investment priority, a figure that conflicts with KPMG's own October 2025 CEO Outlook, which put the number at 71 percent. One case study claimed Emirates uses a mobile chatbot named Sara to converse with passengers and change flights; Sara is in fact a physical robot assistant introduced in 2023 that cannot alter bookings, according to GPTZero's writeup.

KPMG told reporters it expects employees to follow the firm's guidelines on responsible AI use, including human oversight to validate content and verify independent sources. The incident follows a similar withdrawal last month by EY of a loyalty rewards report that contained fake footnotes traced to AI hallucinations.

What this means for job seekers using AI at work

The KPMG episode reframes AI accuracy from an abstract debate into a concrete career risk. A respected global firm shipped a flagship report under its own brand and had to retract it after independent reviewers checked the footnotes. The people who actually drafted, edited and signed off on that document now own that outcome internally. For anyone using ChatGPT, Gemini or Claude in their day job or in a job search, that is the cautionary tale worth memorizing.

Two practical habits follow. First, treat every AI-generated citation as a claim to verify, not a fact to forward. The "vibe citing" pattern GPTZero described is exactly what happens when a model is asked to find examples and rewards plausibility over truth. Open each link. Confirm the publisher, the date and the specific number. If the source does not exist or does not say what the model claimed, cut the line.

Second, use AI to draft and structure, then layer your own judgment on top. The skill that protects your name on a deliverable is the same skill that wins interviews: knowing your material well enough to spot when a tool is wrong. Our guide to job searching in the AI era and our deep work system for junior professionals both make the same point in different contexts. AI is leverage. Unverified AI is liability.

Sources

• "KPMG pulls report on AI usage due to apparent hallucinations" — TechCrunch — https://techcrunch.com/2026/06/13/kpmg-pulls-report-on-ai-usage-due-to-apparent-hallucinations/ — accessed 2026-06-14

• "KPMG's AI report becomes an accidental demo of AI hallucinations" — The Register — https://www.theregister.com/ai-and-ml/2026/06/12/kpmgs-ai-report-turns-into-a-demo-of-ai-hallucinations/5255029 — accessed 2026-06-14

• "Chasing the Hallucinations: KPMG's AI-Powered Attempt at Redefining Excellence" — GPTZero — https://gptzero.me/news/investigations-kpmg/ — accessed 2026-06-14