Top Consulting Group PwC Caught Using AI on “Thought Leadership” Report About AI, Resulting in Corporate Document Filled With Bizarre Hallucinations

PwC, a Big Four consulting giant, published multiple “thought leadership” reports about artificial intelligence that contained numerous AI hallucinations, fabricated citations, and nonexistent frameworks, undermining institutional confidence in the firm’s research quality and raising questions about due diligence across the consulting industry. For institutional investors and enterprise clients relying on consulting firms for strategic AI guidance, the incident exposes gaps in quality control at firms advising on trillion-dollar digital transformation decisions.

• Four PwC reports published between 2024 and 2026 designed to solicit Middle East consulting work contained hallucinated citations, invented frameworks, and contradictory claims flagged by AI detector firm GPTZero.
• One report cited a teenage Medium blogger with 280 followers as a source for a JPMorgan initiative; another repeated the same traffic accident claim three times across two pages using different fabricated sources.
• PwC’s response downplayed findings as “a limited number of supporting citations” needing updates, deflecting accountability rather than disclosing scope or implementing visible remediation.

• 4 PwC reports published between 2024 and 2026 containing AI hallucinations and fabricated research content
• 3 times same traffic accident statistic repeated across two pages with different fabricated sources
• 280 followers on Medium blogger cited as JPMorgan agentic AI initiative source by PwC

PwC, one of the world’s largest professional services firms, published a series of research reports on artificial intelligence and autonomous systems between 2024 and 2026 that were substantially generated or edited with AI tools, but without adequate verification of factual accuracy.

Analysis by GPTZero, an AI detection firm, revealed the reports contained pervasive hallucinations: fabricated academic and industry citations, invented frameworks presented as real-world deployments, and claims that directly contradicted their own cited sources.

The reports, intended to attract consulting engagements in the Middle East, covered high-stakes topics including agentic AI systems, autonomous vehicle governance, and AI integration in investment advisory and compliance, domains where accuracy directly affects client decision-making and regulatory exposure.

PwC’s “Citizen Pulse” Framework Entirely Invented by AI Without Verification

One of the most significant hallucinations involved a framework called “Citizen Pulse,” which a 2025 PwC report presented as an existing government deployment model used globally. GPTZero investigators determined the framework did not exist; it was entirely fabricated by the AI system used to draft or edit the report, complete with supporting citations designed to appear credible.

This represents a critical failure in research methodology: a consulting firm published guidance to government clients about a non-existent policy tool, potentially misleading decision-makers about available technological solutions and best practices.

The Citizen Pulse hallucination is particularly problematic because it exemplifies how AI-generated content can appear plausible to readers lacking deep subject matter expertise. Government officials or enterprise leaders reviewing the report would have encountered what appeared to be a peer-reviewed, consultant-endorsed framework backed by citations, when in fact none of it was real.

This type of fabrication undermines the credibility of consulting advice and creates liability exposure for clients who act on false information presented as research.

Citation Abuse: Medium Blogger With 280 Followers Listed as JPMorgan Initiative Source

Beyond invented frameworks, PwC’s reports contained absurd citation choices that no human researcher would have selected. One report cited a teenage blogger on Medium with 280 followers as a source for a JPMorgan initiative related to agentic AI systems.

The citation, flagged by GPTZero researchers, reveals either a complete failure in fact-checking or reliance on an AI system to generate plausible-sounding references without human verification of their legitimacy or relevance.

A separate report demonstrated even more egregious quality control failures. GPTZero researcher Paul Esau identified a section that repeated the same factual claim, that human error causes 90 percent of traffic accidents, three times within just two pages, each instance citing a different source.

As Esau told the Financial Times, “no human” would construct research this way, suggesting the report was generated or substantially edited by AI without meaningful human review. The repetition with multiple attributions is a hallmark of AI systems that generate citations based on pattern matching rather than actual source consultation.

No human would do this.

Paul Esau, GPTZero researcher

KPMG Similarly Caught Publishing AI-Generated Content on Investment Advisory AI Integration

KPMG, another Big Four firm, faced parallel scrutiny just weeks before the PwC findings became public.

The earlier KPMG case involved a report on integrating AI agents across investment advisory, risk management, and compliance monitoring, domains where institutional investors and asset managers depend on consulting guidance to navigate regulatory and operational risk. Like PwC’s output, the KPMG report was identified as containing substantial AI hallucinations without adequate human verification.

The pattern suggests that reliance on AI-generated “thought leadership” without rigorous fact-checking has become systemic across the consulting industry, particularly for emerging technology topics where consultants may feel pressure to publish quickly and where clients may have limited ability to independently verify technical claims.

The timing and nature of both Big Four incidents indicate a potential industry-wide problem. Consulting firms face competitive pressure to publish research on high-demand topics like AI and autonomous systems, creating incentive to accelerate content production.

Using AI systems to draft or substantially edit reports can appear to reduce time-to-publication, but the cost is accuracy and verifiability. Institutional clients, particularly in regulated industries like finance, insurance, and public sector governance, are now dependent on consulting advice that may be partially or substantially hallucinated.

PwC’s Non-Response Deflects Accountability Rather Than Disclosing Remediation Scope

PwC’s official response minimized the incident and avoided acknowledging the systemic nature of the quality control failure.

In a statement to the Financial Times, PwC Middle East claimed it “takes the accuracy of our published research seriously” and is “updating a limited number of supporting citations” in the reports. The firm also asserted that it maintains “quality control processes for research and content development” that staff are expected to follow.

The language, “limited number of supporting citations”, mischaracterizes findings that included entirely invented frameworks, fabricated sources, and repeated citations with no basis in fact. This framing suggests either PwC does not understand the scope of the hallucinations or is deliberately downplaying the severity to avoid institutional reputational damage.

The response does not indicate whether PwC has retracted the reports, commissioned independent fact-checking audits, or notified clients who may have received consulting recommendations based on the hallucinated research.

It does not specify what changes to quality control processes will prevent recurrence, nor does it acknowledge that multiple reports on AI systems contained hallucinations, revealing a potential problem with how the firm approaches AI-assisted content creation across teams and geographies.

For institutional clients and partners who have relied on PwC guidance based on these reports, the lack of transparency about scope and remediation creates uncertainty about the reliability of other recent research and advisory work.

Institutional Trust Erosion as Consulting Firms Publish Unverified AI-Generated Research

The PwC and KPMG incidents occur within a broader pattern of AI hallucinations compromising institutional credibility across professional services. Lawyers have been sanctioned by judges for submitting fabricated case citations generated by AI systems. Academic journals have been flooded with AI-generated papers containing false claims and invented studies.

Now the Big Four consulting firms, firms hired by governments, large enterprises, and institutional investors to provide expert guidance on complex, high-stakes decisions, are publishing research riddled with fabricated frameworks, nonexistent sources, and contradictory claims.

The stakes are particularly high in the AI domain, where consulting guidance directly influences enterprise investment decisions, government policy, and regulatory strategy. A Fortune 500 company or government agency that commissions a PwC report on AI governance or autonomous systems implementation is paying for expert analysis to reduce decision-making risk. If that analysis is substantially AI-generated without rigorous fact-checking, the firm is not reducing risk, it is obscuring it by providing plaus