Speed is what makes AI useful to a consultant, and it is also the source of the risk. Language models hallucinate: they produce plausible citations, quotes and facts that do not exist, and present them with full confidence. When a tool returns a polished answer in seconds, the temptation to trust it is strong. In most industries that is a quality problem. In consulting, where the product is credibility, it is a business problem, and between 2025 and 2026 it became a documented one.
Key takeaways
- Deloitte, EY and KPMG each published reports in 2025 or 2026 with fabricated citations, quotes or claims. In two of the three cases the research group GPTZero found the errors.
- The mechanism is the same in each case: a trusted firm publishes something false, and its authority discourages readers from checking, so a private mistake becomes a public one.
- The fix is a process rule rather than a tool: no AI-generated claim reaches a client or the public until a human has traced it to a real document.
- Article 4 of the EU AI Act now requires organisations that deploy AI to ensure their staff are AI-literate. This kind of over-reliance is what the article is aimed at.
Why hallucinations are a consulting problem specifically
A consultancy sells two things: an answer, and the confidence to act on it. The second is worth more, and it depends on the client believing that the firm checked. Regulators identified the exposure early. In June 2025 the UK's Financial Reporting Council warned the Big Four's audit teams that they had no performance indicators for AI and asked them to define how the tools affect audit quality.[1] The warning concerned audit, but the underlying concern, firms deploying tools faster than they measure the effect on quality, applies to every practice.
Three documented incidents, 2025 to 2026
The three cases below come from our 2026 report and are limited to what is publicly documented.
| Firm | When | What was published | What was wrong | What happened next |
|---|---|---|---|---|
| Deloitte[2] | 2025 | A report for Australia's Department of Employment and Workplace Relations, published on the department's website | Cited academic research papers that did not exist and quoted a Federal Court judgment that was never made | Deloitte disclosed it had used Azure OpenAI to help produce the report and, in October, agreed to refund the final instalment of a roughly $290,000 contract |
| EY[3] | May 2026 | A study on loyalty rewards programmes, used by EY consultants to market the firm's cyber-security work in Canada | GPTZero found more than half a dozen hallucinated footnotes pointing to pages that did not exist or did not contain the cited information, including a reference to a McKinsey report that does not exist | EY withdrew the study |
| KPMG[4] | Published Oct 2025, removed June 2026 | “Redefining excellence in the age of agentic AI” | Claimed UBS, NHS Greater Manchester, Swiss Federal Railways and Transport for London were running AI agents in ways those organisations said they were not | GPTZero identified the inaccuracies; KPMG removed the report from its website |
Deloitte confirmed it used Azure OpenAI on the Australian report. The EY and KPMG errors were identified and published by GPTZero, a research group that specialises in detecting AI-generated content. We did not re-verify each footnote; the facts above are as those sources reported them.
The pattern: authority turns a private error into a public one
The three cases share one pattern. A firm whose authority leads people to trust its work published claims that were not true, and nobody checked because the firm's name served as the check. The more credible the publisher, the less likely anyone is to verify, and the further a fabricated fact travels before someone does.
“Publishing a report online is essentially a form of data injection into the pool of knowledge that is the internet. When the report includes fake information (either vibed citations or false claims) it can ‘poison the well’ by misleading future researchers, especially if the report is published by a well-known consulting firm and hosted on a high-traffic website.”[3]
How errors propagate once published
The GPTZero authors' phrase, poison the well, describes a second-order effect that consulting firms have not previously had to consider. A Big Four report is read many times: it is cited in client decks, quoted in tenders, summarised by journalists and, increasingly, ingested by the AI systems that will answer the next analyst's question. A McKinsey report that never existed, cited in an EY footnote, does not stay in the footnote. It becomes accepted as fact because a trusted firm cited it.
The costs are therefore asymmetric. Skipping verification saves minutes per report. One public correction leads every future client to apply a discount to everything the firm says. Consulting firms have always charged for their credibility, and an unverified footnote spends it.
Building verification into the delivery process
The remedy is a process rule rather than a change of tools: no AI-generated claim reaches a client or the public until a human has traced it back to a real document. The table below shows where that rule fits in a delivery process.
Where verification belongs
| Stage | What to check | Who owns it |
|---|---|---|
| Research | Every source an AI tool surfaces is opened and read rather than listed. Sources that cannot be located are removed rather than paraphrased. | The consultant who ran the query |
| Drafting | Quotes, figures and named examples are tagged as verified or unverified in the draft itself, so nothing unverified can be mistaken for fact at review. | The drafter |
| Review | A reviewer samples citations against the underlying documents, with a mandatory full check for anything that will be published externally. | Engagement manager or partner |
| Publication | A final pass specifically for fabricated references, quotations and claims about named organisations, the failure modes in the 2025 to 2026 incidents. | A named owner, recorded |
Two design choices matter most. First, make verification a named step with a named owner: the firms in the table above have general quality standards, and general standards are what failed. Second, treat external publication as a firm threshold. An internal note can carry an unverified claim with a flag on it; a report on a government website cannot.
AI literacy, and what the EU now requires
Regulation points the same way. Article 4 of the EU AI Act requires organisations that deploy AI to ensure their staff have adequate AI literacy,[5] and the failure it targets is the one the three cases share: people trusting output they had no basis to trust. Literacy in this sense means knowing that the model will invent a citation rather than admit it has none, and checking as a matter of habit. That habit can be trained.
The cases and the process above are developed in chapter 5[6] of The State of AI in Consulting 2026. For the regulatory detail, see the EU AI Act for consulting firms.
Notes and sources
- UK Financial Reporting Council, communication to Big Four audit teams on AI performance indicators, June 2025
- Australian Department of Employment and Workplace Relations, Deloitte report, disclosure of Azure OpenAI use and partial refund, 2025
- GPTZero (Om Ogale, Paul Esau, Alex Cui), analysis of EY loyalty-rewards study, May 2026
- GPTZero, analysis of KPMG “Redefining excellence in the age of agentic AI” (October 2025), report removed June 2026
- Regulation (EU) 2024/1689 (EU AI Act), Article 4 on AI literacy
- Spaik, The State of AI in Consulting 2026, chapter 5
Figures attributed to third parties are their own reported data; Spaik did not produce those statistics. Where the text offers an interpretation, it is Spaik's own.
Continue reading
Working with Spaik
Verification is a skill. It can be trained.
Every Spaik programme builds source verification and critical thinking into the exercises, on the tools your teams already use. Consultants leave knowing where the model can be trusted, where it cannot, and what has to be traced to a real document before a client sees it.