GWI finds 58% of surveyed marketing leaders acted on wrong or misleading AI insights

Wrong or misleading AI insights have already shaped significant decisions for 58% of senior marketing, insights and agency professionals surveyed by GWI. The company released Artificial Confidence: The AI Trust Gap on November 10, 2026, finding that reported consequences extend beyond correcting work to customer complaints, financial losses and reputational damage.

GWI surveyed 1,883 professionals in July 2026: 850 in the United States, 665 in the United Kingdom and 368 in Singapore. Respondents worked at companies with at least 500 employees or £20 million in annual turnover. The sample therefore describes a particular group of senior professionals, rather than business leaders across every occupation, company size or country.

The findings were published in GWI’s press release, distributed through GlobeNewswire and carried by Yahoo Finance as a paid release. They are respondents’ accounts of their experiences, not independently audited business outcomes.

Errors show up in everyday operations

Across the sample, 82% reported at least one negative business consequence from inaccurate AI output. The consequences ranged from operational friction to damage outside the organization. Financial losses were reported by 27%, reputational damage by 26% and weaker campaign performance by 21%.

38%Internal rework

Respondents who reported needing internal rework because of inaccurate AI output.

36%Delayed decisions

Respondents who reported delays to decisions because of inaccurate AI output.

34%Customer complaints

Respondents who reported customer complaints resulting from inaccurate AI output.

GWI argues that these costs can disappear into ordinary operating work rather than being traced to the original flawed insight. Its warning is about how businesses account for AI’s value: efficiency gains may be counted while time spent fixing errors is absorbed elsewhere.

Checking AI with another AI

Awareness of the risk was high: 85% rated decisions based on inaccurate AI insights as a significant risk. Routine verification was less common. Only 36% said they verified AI-generated insights “most of the time,” while 22% checked only occasionally. Separately, 36% said they cross-checked using a different AI tool. That answer describes a checking method, not evidence that the second tool corrected the first.

The verification figures also vary by country. GWI puts the share checking most of the time at 43% in the US and 29% in the UK; the overall figure is 36%. ITPro’s coverage focuses on UK findings, including a 56% share reporting a major decision based on a wrong or misleading AI insight.

Outdated information is another reported problem. In the three-country survey, 63% said they had unknowingly acted on stale data or insights. Meanwhile, 72% agreed that AI interfaces can make old information appear more relevant than it is. GWI founder and CEO Tom Smith argues that organizations need to ground AI-informed decisions in fresh, representative evidence from real people, as ITPro reported.

Confidence does not mean replacing research

Respondents were not uniformly skeptical of AI’s ability to explain consumers. Eighty-four percent felt confident that AI could accurately explain why people behave as they do. Yet 80% also said understanding motivations still requires human research, and 75% said AI would never solely replace direct consumer research.

GWI has a commercial interest in that distinction: it describes its business as providing companies and AI systems with proprietary data about people. Its recommendation to ground AI in human research aligns with that offering.