KPMG Pulls Agentic AI Report Over Apparent Hallucinations: What Happened?

Primary source: TechCrunch
Original article: https://techcrunch.com/2026/06/13/kpmg-pulls-report-on-ai-usage-due-to-apparent-hallucinations/
Topic: KPMG, agentic AI, AI hallucinations, source verification, consulting reports, responsible AI usage
Original article date: June 13, 2026
Prepared: June 14, 2026
Scope: news analysis, not legal advice

Quick summary

TechCrunch reports that KPMG pulled a report titled “Redefining excellence in the age of agentic AI” after several organizations named in the report said its claims about their AI usage were untrue or misleading.

According to TechCrunch, research group GPTZero identified multiple inaccuracies in the report, which had been published in October 2025. GPTZero told the Financial Times that the inaccuracies appeared to stem from AI hallucinations. In other words, a professional services firm appears to have used AI to help write a report about AI, and the result contained claims, citations, or case studies that were not adequately verified.

The important point is not only that KPMG removed the report. The larger issue is that if major consulting firms publish AI-generated thought leadership without rigorous verification, false information can spread into journalism, blogs, enterprise decisions, and even future AI model training data.

What happened?

TechCrunch says KPMG removed the report after numerous organizations disputed claims about how they were using AI. Organizations mentioned include UBS, the U.K.’s National Health Service, Swiss Federal Railways, and Transport for London. According to TechCrunch, these organizations told the Financial Times that the report’s claims were untrue or misleading.

KPMG said it removed the report from its websites while conducting its own investigation. A KPMG spokesperson also said the firm expects its people to follow responsible AI guidelines, including human oversight to validate content and verify independent sources.

In short:

What does “AI hallucination” mean in this case?

An AI hallucination is content generated by AI that sounds plausible but is false, unsupported, wrongly sourced, or misinterpreted.

In a professional report, hallucination can appear as:

a citation to a report that does not exist
a real source used to support a claim it does not actually make
a case study attributed to an organization that did not do it
several real sources blended into one false conclusion
a source title paraphrased so heavily that it cannot be verified
one narrow example inflated into a broad trend

The dangerous part is that these errors often look professional. If readers do not check every source, they may trust the claims because they come from a well-known brand.

What was allegedly wrong with the KPMG report?

TechCrunch summarizes that the report made claims about how certain organizations used AI, but those organizations said the claims were untrue or misleading. The Financial Times reported more specifically that the report exaggerated AI adoption by UBS, the NHS, Swiss Federal Railways, and Transport for London.

TechRadar, citing GPTZero’s analysis, reported that the KPMG document contained 45 citations, but only five accurately pointed to real sources; the rest were described as fabricated, distorted, or misleading. GPTZero used the term “vibe citing” to describe plausible-looking citations that fail verification.

This should still be described carefully. The claims come from reporting by TechCrunch, FT, TechRadar, and GPTZero; KPMG has removed the report and is investigating, so the final institutional conclusion depends on that review and any follow-up from the organizations involved.

Why is this serious?

KPMG is not a small blog. It is one of the world’s largest professional services firms. Reports from major consultancies are often cited by businesses, investors, journalists, and policymakers.

If such a report contains false information, the damage is not limited to one PDF. It can spread through a chain:

This is the risk of “second-hand hallucination”: AI-generated misinformation gets reused until it starts looking like common knowledge.

Why is this especially ironic?

The pulled report was about agentic AI — AI systems that can perform multi-step tasks, use tools, and operate more autonomously than ordinary chatbots.

Consulting firms are selling services that help companies adopt AI responsibly. So when a major consulting firm’s AI report appears to contain hallucinations, the message that “AI creates value when governed properly” becomes weaker.

Plainly:

If an AI advisory firm cannot verify its own AI report,
clients have reason to question its AI governance processes.

What is “vibe citing”?

“Vibe citing” is a sarcastic label for citations that feel credible but are not actually correct.

Examples include:

The problem is that citation formatting creates trust. Readers see footnotes, organization names, dates, and technical language, and assume the content has been verified.

Lesson for AI report writers

The key lesson is simple: AI may assist writing, but it must not replace source verification.

A minimum verification workflow should include:

For enterprise-facing content, especially reports used to market consulting services, the verification standard should be even higher.

Lesson for companies reading consulting reports

Companies should not trust a report only because a major consulting logo appears on it.

When reading AI reports, check:

Readers do not need to be cynical about every report. But important claims should be verified before they are used to make decisions.

Lesson for startups and AI content creators

If you use AI to write blogs, whitepapers, SEO articles, or market reports, the same risk applies.

AI is good at producing fluent language. But when it needs sources, numbers, company names, report names, dates, or case studies, it may generate false content.

Practical checklist:

For news and SEO/GEO content, source hygiene matters because false information can damage long-term site credibility.

Why does this happen in thought leadership?

Thought leadership is speed-driven. The AI market moves quickly, and companies want to publish early to shape the conversation and attract clients.

This creates three risks:

The longer the report, the more case studies and footnotes it contains, the higher the risk if the verification process is weak.

Comparison with the EY case

TechCrunch notes that EY also withdrew a report on loyalty rewards programs last month after it appeared to contain fake footnotes and AI hallucinations. The Financial Times reported that EY Canada’s study contained made-up data, misattributed citations, and a reference to a McKinsey report that did not exist.

This suggests the issue is not isolated to one firm. It is a broader problem in professional services: consulting firms are using AI to produce more content, but verification workflows have not always kept up.

Impact on trust in agentic AI

Agentic AI is already a sensitive topic because it promises more autonomous systems that can act over multiple steps. If reports praising agentic AI contain false examples, readers may react skeptically.

Instead of believing that agentic AI is widely adopted and effective, they may ask:

how many case studies are real?
how many numbers were verified?
who is accountable when AI writes something false?
are consulting firms overstating AI adoption to sell services?

This can reduce enterprise trust in AI unless major firms become more transparent about how their content is created and verified.

Analysis

KPMG’s withdrawn report is a useful example of today’s AI paradox: the organizations advising enterprises on responsible AI must also prove that they themselves use AI responsibly.

AI is not the only problem. The deeper issue is process. If AI is used for drafting and humans verify rigorously, risk can be reduced. If AI is used to produce content quickly and citations are checked superficially, the report becomes a misinformation vector.

Conclusion

KPMG’s withdrawal shows that in the AI era, references are not a side detail. Citations, footnotes, case studies, and numbers are core parts of the content product.

Short version:

Without that, a report about AI can become one of the clearest examples of AI’s own risks.

SEO title suggestions

• KPMG Pulls Agentic AI Report Over Apparent AI Hallucinations: What Happened?
• What Is “Vibe Citing” and Why Did KPMG’s AI Report Spark Controversy?
• When an AI Consulting Report Contains AI Hallucinations: Lessons From KPMG
• KPMG, GPTZero, and the Risk of Fake Citations in AI Reports

SEO meta description

TechCrunch reports that KPMG pulled its “Redefining excellence in the age of agentic AI” report after several organizations disputed claims about their AI usage. This article explains AI hallucinations, vibe citing, fake citations, impact on trust in agentic AI, and lessons for enterprises, startups, and AI content creators.

Keywords

KPMG, agentic AI, AI hallucination, GPTZero, vibe citing, fake citations, AI report, TechCrunch, Financial Times, EY hallucination, AI governance, responsible AI

References

1. TechCrunch — KPMG pulls report on AI usage due to apparent hallucinations: https://techcrunch.com/2026/06/13/kpmg-pulls-report-on-ai-usage-due-to-apparent-hallucinations/
2. Financial Times — KPMG report contained AI hallucinations on benefits of AI: https://www.ft.com/content/b3828e92-4961-4b39-84f0-c42f33be3c3f
3. TechRadar — A major KPMG report on AI was found to be chock-full of AI hallucinations: https://www.techradar.com/pro/a-major-kpmg-report-on-ai-was-found-to-be-chock-full-of-ai-hallucinations
4. Financial Times — EY retracts study after researchers discover AI hallucinations: https://www.ft.com/content/a61cbcae-95e4-4449-86e1-ef40fb306f4e