A recent investigation has revealed that a significant majority of references within a flagship report by professional services giant KPMG were found to be inaccurate or entirely fabricated. The report, titled ‘Total Experience: Redefining Excellence in the Age of Agentic AI’ and published in October 2025, formed part of KPMG's annual global customer experience excellence study. The findings underscore a growing concern about the reliability of AI-generated content within the professional services sector.
The probe, conducted by AI detection software GPTZero, meticulously examined the 45 citations presented in KPMG's report. It concluded that a staggering 40 of these citations were fake, failing to point to real, uncorrupted sources. GPTZero’s analysis also highlighted that roughly half of the factual claims supported by these citations were either false or misattributed. For instance, the report cited a 2019 East Japan Railway press release as evidence of AI agent usage, despite the term 'agentic AI' only entering public discourse in 2024.
KPMG itself has acknowledged the phenomenon, coining the term 'vibe citing' to describe how generative AI tools can inadvertently create fake references, blend real sources, or heavily paraphrase titles. The investigation also found instances where KPMG's large language model (LLM) confused article subjects with their authors, such as crediting Transport for London (TfL) as the author of a blog post about the organisation, rather than the actual blogger. Furthermore, the report contained internal contradictions, citing 'KPMG research' that claimed 55 per cent of CEOs prioritised AI investment, while the firm’s own 2025 CEO Outlook, published in the same month, stated this figure was 71 per cent.
The implications of these errors extend beyond KPMG, as the flawed statistics and claims from the report have already been recycled by various industry publications and even a Czech newspaper. Worryingly, these inaccuracies are now being cited directly by other LLMs like ChatGPT and Gemini, potentially propagating misinformation further across the digital landscape. Following these revelations, KPMG has removed the 'Total Experience' report from its homepage.
A spokesperson for KPMG informed the Financial Times that the firm “takes the accuracy and integrity of its published content seriously” and confirmed the report's removal while an investigation into its publication is underway. The firm reiterated its expectation for all staff to adhere to guidelines on the responsible use of AI, emphasising the necessity of human oversight for content validation and source verification. The author of the report, Paul Esau, suggested in the findings that “no human at KPMG double-checked the citations, the claims, or the sources before Total Experience was published.”
This incident is not isolated, reflecting a broader challenge within professional services. Last October, Deloitte issued a partial refund to the Australian federal government due to AI-induced errors in a report. Similarly, London-headquartered law firm Pinsent Masons faced criticism from a High Court judge after a lawyer submitted AI-generated letters containing false legal information. In April, US law firm Sullivan & Cromwell apologised to a judge for AI-generated 'hallucinations' in a high-profile case filing, and in May, EY withdrew a study on loyalty rewards programmes that also contained apparent AI hallucinations and fake footnotes.
Source: GPTZero, Financial Times