While some sectors and companies are still figuring out how to use AI responsibly, two of the most trusted names in professional services just provided a costly lesson in what happens when no one checks AI-created content marketing or technical writing.
In June 2026, KPMG pulled a report titled"Redefining Excellence in the Age of Agentic AI" after theAI-detection firm GPTZero found that 40 of its 45 citations were fabricated —only five pointed to real, intact sources. Organizations named in the report, including UBS, the UK's National Health Service, Swiss Federal Railways, andTransport for London, told the Financial Times that its claims about their ownAI usage were untrue or misleading. KPMG's statement on the episode said the quiet part out loud: "We expect all our people to follow our guidelines on the responsible use of AI, including human oversight to validate content and verify independent sources." (TechCrunch)
A month earlier, EY Canada quietly removed a study called "Points of Attack: Uncovering Cyber Threats and Fraud inLoyalty Systems" after GPTZero found that 16 of its 27 references were hallucinated — including a fabricated citation behind the report's headline$200 million claim. EY consultants had been actively using the study to pitch cybersecurity services to clients. (International Accounting Bulletin)
These two examples are prominent, but they are hardly isolated, and they aren't confined to professional services. We hear the same story directly from colleagues at technology companies that have been pushed to use AI for content marketing. What we hear, consistently, is that the output is rarely a free lunch: it's often inaccurate, frequently off-brand, and routinely creates more fact-checking and rewriting work than writing it correctly the first time would have.
Technical Content Raises the Stakes
The problem compounds in technical writing for sectors such as healthcare, technology, and utilities where the underlying information is genuinely complex. First of all it's scattered across product specs, code, compliance frameworks, and the heads of a handful of subject-matter experts, and it requires real understanding of how a product or operation actually works, not just how it's described on the public web. A general-purpose AI model has no access to your architecture, your latest release notes, or the edge cases only your engineers are aware of. When AI fills those gaps, it doesn't flag the guess as a guess, it states it as fact, with the same confident tone as everything else. In a blog post about marketing trends, that might mean an awkward correction. In a technical datasheet, an integration guide, or a whitepaper, it can mean a customer building on a capability that doesn't exist, a regulator reading a compliance claim that isn't true, or a partner making a decision based on a number nobody verified.
The Cost of AI is Bigger than theCorrection
The incentive to lean on AI for content is obvious. Managers are being pushed to lower cost and deliver faster turnaround.But KPMG and EY are firms whose entire business is built on accuracy and independent verification and even they got burned, publicly, by skipping the human-verification step. The damage in cases like this rarely stops at a quiet edit. It shows up as press coverage that outlasts the original content by months, as named third parties publicly disputing your claims, and as clients and partners quietly re-checking everything else you've published. It also shows up inside the building in the form of the internal furor of teams scrambling to issue corrections, subject-matter experts who feel blindsided by claims attributed to them, and a marketing or comms function that loses credibility with the very engineers and leaders it depends on for accurate information next time. Once a colleague or client has caught one fabricated detail, every other claim in the document, and the next one, gets read with suspicion. That's an expensive trust deficit to rebuild, and it's exactly the opposite of what content marketing is supposed to build in the first place.
We Keep Humans at the Center
This is why StoryAZ treats AI as a tool a skilled writer uses, never as a replacement for one. Every project starts with a writer, editor, or researcher who understands the subject, talks to the people who actually build or operate it, and checks every claim against a primary source. AI can help us brainstorm angles, summarize background reading, or accelerate a first pass at structure. It does not get to originate a fact that ends up in a client's deliverable unverified, and it does not get the final word on brand voice, nuance, or judgment calls that require actually understanding the audience.
Clients come to StoryAZ for technical writing specifically because getting it right matters. When content carries your company's name, there's no acceptable error rate on the things that are checkable. Speed and cost matter but not more than your own personal reputation. AI is never a substitute for the human research, domain expertise, and editorial judgment that keep your content accurate, on-brand, and worth the trust your audience puts in it.