In 2026 two of the world’s largest professional-services firms withdrew marketing reports because the AI that helped write them invented the evidence. Trust has become the scarcest asset in the category.
What happened
KPMG withdrew its “Total Experience” report after an investigation found that 40 of its 45 citations were fabricated — references to sources that did not exist. Within weeks, EY pulled a 2026 study on loyalty programmes for similar reasons. In both cases the likely culprit was an AI referencing tool left unchecked.
The episodes were widely reported — by The Register and City AM, and logged in the OECD’s AI incidents database — and they were not isolated. They were a pattern.
The uncomfortable lesson
Generative tools scale content and scale error with equal ease. Authority no longer follows volume; it follows verifiability. An insight is only as good as the evidence beneath it, and that evidence now has to be checked by someone who can be held to account for it.
For the buyer of advice, “who stands behind this number?” is suddenly a strategic question.
A standard, not a footnote
We treat zero invention as a standard, not a compliance note: every figure, source and case is verified before it is published. In a market where the largest names have been caught out, that is not housekeeping. It is positioning.
Common questions
Why did KPMG and EY withdraw reports in 2026?
Both contained AI-fabricated citations — KPMG’s report had 40 of 45 references invented — and were withdrawn after public scrutiny.
How do you avoid AI hallucinations in content?
Verify every figure, source and case before publishing, and treat zero invention as a standard rather than an afterthought.