{"slug": "vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look", "title": "Vibe citing: how KPMG used AI to write a report about AI and AI made them look like fools", "summary": "KPMG, a global consulting firm with 250,000 employees, published an AI-generated report in October 2025 titled 'Total Experience: Redefining Excellence in the Age of Agentic AI.' An audit by GPTZero found that 40 of 45 citations were invented, yielding an 11% accuracy rate, and half of the factual claims were false or misattributed. The Financial Times confirmed that companies cited as success stories, including UBS and NHS United Kingdom, had no such AI agent deployments, leading to the coining of the term 'vibe citing' for AI-generated fake references.", "body_md": "*by t474-r0b07*\n\nThere are companies that charge you to tell you how to use AI responsibly.\n\nKPMG is one of them.\n\n250,000 employees. 138 countries. Decades advising governments and corporations on how to avoid costly mistakes.\n\nIn October 2025 they published a report titled *\"Total Experience: Redefining Excellence in the Age of Agentic AI\"*.\n\nThey wrote it with AI.\n\nThe AI invented 88% of the sources.\n\nNobody verified anything.\n\nThey published it anyway.\n\nAn agentic AI is not a chatbot.\n\nNot the assistant that answers your questions. It's a system that makes decisions and executes actions on its own, without a human approving each step. You give it an objective and it acts, corrects, moves forward.\n\nIt's the product everyone in the tech sector was selling in 2025.\n\nKPMG was selling it too.\n\nThat's why they needed a report proving their clients were already using it.\n\nSpoiler: they weren't. And the report invented it anyway.\n\nGPTZero — a company specialized in detecting AI-generated content — ran a full audit on the report.\n\nFirst: what is an AI hallucination, because the term is going to come up a lot.\n\nWhen a language model doesn't have the information you ask for, it doesn't say \"I don't know.\" It generates a response that *sounds* correct. It invents with the same confidence it would use if it actually knew the truth. Perfect format. False content. No warning.\n\nThat's a hallucination.\n\nNow the numbers from the KPMG report:\n\n```\nTOTAL CITATIONS:      45\nREAL CITATIONS:        5\nINVENTED CITATIONS:   40\nACCURACY RATE:      11.1%\n```\n\n40 of 45 citations have invented titles, authors that don't exist, or sources that don't say what KPMG claimed they said.\n\nHalf of the factual claims in the report are false or misattributed.\n\nA firm that charges for intellectual rigor published a document with 11% accuracy.\n\nThe Financial Times contacted the companies listed as success stories.\n\n**UBS** — false.\n\n**NHS United Kingdom** — false or misleading.\n\n**Swiss Federal Railways** — false.\n\n**Transport for London** — \"misleading.\"\n\nTransport for London said the claims that they were using AI agents to predict congestion and coordinate the network were *misleading*.\n\nNHS Greater Manchester said the description of using agentic AI to organize patient records and predict hospital readmissions *\"doesn't really align\"* with reality.\n\nKPMG put their logos on fiction without asking permission.\n\nAnd billed them as success stories.\n\nThe model was instructed to find cases of companies using agentic AI.\n\nIt didn't find enough — because in many sectors they simply don't exist yet.\n\nSo it did the most comfortable thing: it generated them.\n\nIt cited a **East Japan Railway press release from 2019** as evidence of agentic AI adoption.\n\nThe term *agentic AI* didn't exist in public discourse until 2024.\n\nThe model traveled five years back in time, reformulated an unrelated document, and presented it as proof of something that hadn't happened yet.\n\nIt wasn't an error. It was the easiest answer to the prompt.\n\nThe model doesn't understand the difference between inventing and remembering. It generates what fits. If it doesn't exist, it builds it. And it does so with the same fluency it would use to cite something real.\n\nGPTZero coined the term: **vibe citing**.\n\nTo understand it you first need to understand *vibe coding* — writing code without understanding what it does. You ask an AI to generate the code, you copy it, it kind of works, and you move on without reading a line. The vibe is right. The understanding, zero.\n\nVibe citing is the same thing but with bibliography.\n\nThe model generates references that *sound* academic because it processed millions of papers. The structure is correct. The doi has the exact format. The year is right.\n\nThe content is fiction.\n\nAnd the world's largest firm in responsible AI consulting didn't verify a single one before publishing.\n\n``` python\ndef verify_sources(citations):\n    # TODO: implement before publishing\n    pass\n\npublish_report()  # called without verifying anything\n```\n\nThis is not a technical error.\n\nIt's a process decision. Or the absence of one.\n\nThere's a detail that turns negligence into something almost poetic.\n\nThe report cites \"KPMG research\" claiming that **55% of CEOs prioritize AI as their main investment**.\n\nThe *KPMG 2025 CEO Outlook* — published the same month, by the same company — says **71%**.\n\nThe model didn't just invent external sources.\n\nIt invented data from the company that was using it and contradicted it with that same company's real data from the same period.\n\nKPMG cited KPMG incorrectly in a KPMG report.\n\nPage 42.\n\nKPMG claims that Emirates adopted a mobile chatbot called **Sara** that can converse with passengers and change their flights.\n\nReality:\n\nThree claims. None correct on what matters.\n\nThe model took real information about Sara, reformulated it to fit the narrative it needed, and presented it as an agentic AI success story.\n\nThis is not a writing error. It's construction of fiction using real data as scaffolding.\n\nThis is where it stops being an isolated corporate scandal.\n\nGPTZero has been documenting the same pattern for months:\n\nThree of the Big Four in consecutive months.\n\nAll selling responsible AI consulting.\n\nAll publishing hallucinations as research.\n\nThe pattern isn't coincidence. It's market pressure: the client wants the report, the report needs data, the data doesn't exist yet, the model generates it, nobody verifies because verification takes time and the client already paid.\n\nAI is not the problem.\n\nThe economic incentive to appear to know more than you do — that's the problem.\n\nHere's the data point almost no media outlet is discussing.\n\n**The false statistics from the KPMG report are already being reproduced by ChatGPT and Gemini.**\n\nI need to explain why that's structurally serious and not just anecdotal.\n\nFor months the report was published on KPMG's domains. The crawlers that feed language models index sources by authority. KPMG has maximum authority: global company, old domain, millions of visits, decades of institutional credibility.\n\nThe models ingested that content as verified truth.\n\nNow when someone asks ChatGPT or Gemini about agentic AI adoption, they can return the false data from the report — not as \"I found this at KPMG\" but as their own knowledge, without attribution, without warning.\n\nThe full cycle:\n\n```\nmodel hallucinated data\n    → KPMG published without verifying\n        → crawlers indexed it as high-authority source\n            → other models ingested it as truth\n                → user receives the original hallucination as fact\n```\n\n**high-authority source + false data + model ingestion = untraceable disinformation.**\n\nYou can't trace the origin. You can't disinfect the source. The error already lives inside the models you consult every day.\n\nAnd the report has already been retracted. But the data keeps circulating.\n\nTaking down the PDF didn't deindex anything.\n\nKPMG's spokesperson declared after withdrawing the report:\n\n*\"We expect all our staff to follow our guidelines on responsible AI use, including human oversight to validate content and verify independent sources.\"*\n\nTranslation: we have guidelines. Someone didn't follow them. We're investigating.\n\nWhat they didn't say: how a flagship report on responsible AI, with the KPMG logo, published on their official channels, passed through their entire internal review process without anyone verifying a single one of the 45 citations.\n\n250,000 employees.\n\n5 valid citations.\n\nNobody asked anything.\n\nModels do exactly what they were designed to do: generate coherent and plausible text based on learned patterns.\n\nThey don't lie. They have no concept of lying. They generate what fits.\n\nThe problem is human: using AI as a researcher without a verification loop isn't efficiency. It's delegating truth to a system that has no concept of truth, and signing your name on top.\n\nKPMG didn't build a report with AI.\n\nThey built the appearance of a report and sold it as research.\n\nThe difference isn't semantic.\n\nIt's the difference between knowing something and appearing to know it.\n\nIn 2025, the world's largest firms chose to appear.\n\n**primary sources — verify yourself:**\n\n*t474-r0b07 — Tarija, Bolivia*\n\n*github.com/t474-r0b07*", "url": "https://wpnews.pro/news/vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look", "canonical_source": "https://dev.to/t474r0b07/vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look-like-fools-52o5", "published_at": "2026-06-17 03:40:21+00:00", "updated_at": "2026-06-17 03:51:16.449059+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-safety", "ai-ethics", "ai-agents"], "entities": ["KPMG", "GPTZero", "Financial Times", "UBS", "NHS United Kingdom", "Swiss Federal Railways", "Transport for London", "East Japan Railway"], "alternates": {"html": "https://wpnews.pro/news/vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look", "markdown": "https://wpnews.pro/news/vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look.md", "text": "https://wpnews.pro/news/vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look.txt", "jsonld": "https://wpnews.pro/news/vibe-citing-how-kpmg-used-ai-to-write-a-report-about-ai-and-ai-made-them-look.jsonld"}}