GWI survey: 82% of marketing leaders report harm from wrong AI insights

Eighty-two percent of marketing leaders have felt at least one adverse outcome from inaccurate AI insights, according to a GWI survey that [Marketing Dive reported](https://www.marketingdive.com/news/ai-derived-audience-insights-hurt-results-what-the-numbers-say/874565/) on 13 November 2026. The most common were internal rework (38%), delayed decisions (36%) and customer complaints (34%). Financial loss was reported by 27%, reputational damage by 26% and weaker campaign performance by 21%.

The report, "The AI Trust Gap", draws on 1,883 senior insights, agency and marketing professionals in the UK, the US and Singapore, surveyed in the third quarter of 2026. All worked at companies with more than 500 employees or £20 million ($26.5 million) in annual turnover.

Heavy use, light checking

• Nearly 90% use AI daily, and 80% say they frequently encounter inaccurate or misleading AI findings.
• 36% say they verify AI insights "most of the time".
• 63% have used an insight they later realised was out of date, and 68% think colleagues do the same without noticing.
• Over half of senior decision-makers have made a significant business decision on AI insights that turned out to be incorrect or misleading.

Audience profiles are where it shows

Nearly 70% of respondents have seen AI fail to reflect their organisation's target audience in a profile, and 76% think it can reinforce outdated assumptions. Yet 81% feel confident using AI for audience profiles and segmentation (sorting customers into groups to target), and 84% believe it can accurately explain human behaviour. Seventy-one percent say AI can tell you "what" but not "why".

Jason Mander, GWI's chief insight officer, told Marketing Dive: "We've got somewhat lost in the buzz and the excitement, and haven't been as rigorous as we otherwise would be." He added that companies generally "haven't put the type of rigorous frameworks in place that you would expect with a new type of technology".

How much weight the numbers bear

• Everything is self-reported. No campaign results or accounts were examined, and no loss is sized: "financial loss" could be a wasted afternoon or a wasted quarter.
• There is no baseline. The survey does not say how often insights from agencies, panels or analysts turned out wrong before AI.
• The 82% is a share of "marketing leaders". Marketing Dive does not say how many of the 1,883 that is, or how respondents were recruited.
• Marketing Dive does not describe GWI's business or say who paid for the survey. Find out before leaning on the finding that 75% believe AI will never replace direct human research.
• The people side is missing: the report as covered does not say who does the checking where it happens, or how long it takes.

What to do if you run marketing or marketing operations

• Pick the three places AI-derived insight reaches spend: audience profiles, segment definitions and campaign briefs. For each, name one person who confirms the claim against a dated first-party or research source before it is approved.
• Require a date and a source on every AI-generated audience statistic in a brief. Out-of-date insight was the failure 63% admitted to.
• Start a log this quarter of rework, complaints and campaign misses traced to a wrong AI output, with hours and money against each, so you have a figure of your own where the survey has none.