AI is already influencing marketing decisions at scale, but a new GWI report puts a cost on trusting its answers too quickly. In research released November 10, 58% of surveyed senior marketing, insights and agency professionals said they had made a significant business decision based on an AI-generated insight that later proved wrong or misleading. The finding matters because an inaccurate audience assumption can travel from a prompt into a campaign brief, a media plan and customer-facing work before anyone checks the evidence.

The signal is not that businesses should stop using AI. It is that speed without traceability can create a quiet form of decision debt: the rework, missed timing and customer damage that may be booked as ordinary operating friction rather than traced to a flawed insight.

The AI marketing risk is bigger than hallucinations

GWI reports that 88% of respondents use an AI tool daily, while 80% frequently encounter AI findings they consider inaccurate, misleading or untrue. The gap between routine use and routine review is notable: just 36% say they verify AI-generated insights “most of the time,” and another 36% say they cross-check with a different AI tool. A second model can produce a second answer, but it is not independent evidence.

The reported consequences reach beyond embarrassing factual errors. Among respondents, 38% cite internal rework, 36% delayed decisions and 34% customer complaints; 27% report financial loss, 26% reputational damage and 21% weaker campaign performance. These outcomes can overlap, so they should not be added into one total. GWI’s founder and CEO Tom Smith captured the operational problem: “AI sounds just as certain when it’s wrong as when it’s right.”

For marketers, the exposure is clearest when AI supplies a claim about people: who is buying, what motivates them, which message will resonate, or why performance changed. A polished answer can make an old or unrepresentative assumption look like fresh customer research. The full GWI report, Artificial Confidence: The AI Trust Gap, also finds that 63% of respondents had unknowingly used outdated data or insights.

Read the percentages as a warning, not a universal failure rate

The study surveyed 1,883 senior marketing, insights and agency professionals in the UK, United States and Singapore. Respondents worked in qualifying sectors and at companies with at least 500 employees or £20 million in annual turnover. That makes the findings relevant to larger organizations and agency teams, but it does not make the sample representative of every business, small-business owner or AI use case.

There is also a transparency caveat: GWI says the survey was conducted in Q2 2026, while its report page leaves the exact fieldwork dates and survey method “to be confirmed.” The numbers are respondents’ reported experiences, not an independent audit proving AI caused each business outcome. Treat them as evidence of a governance and measurement problem worth testing in your own operation, not as a forecast that a set share of your campaigns will fail. The GWI release provides the headline figures; the report details its sample and caveats.

Build verification into the decision, not after it

Businesses do not need a manual fact-check on every brainstorm. They do need controls that scale with the consequence of acting on an answer. Use a simple three-tier rule:

  • Explore: For low-stakes ideation, let teams use AI freely, but label assumptions and do not present generated claims as customer evidence.
  • Validate: Before audience profiles, campaign briefs or budget recommendations move forward, require a named owner to inspect the underlying source, date, geography, sample and method. If the output makes a customer claim, compare it with first-party behavior or credible research about real people.
  • Approve: For significant spend, sensitive targeting or reputational risk, require a second-person review and a measured pilot with a clear success metric and stop condition.

Make the model show its work, but verify the cited source rather than trusting the citation text. Record when the evidence was collected and when it must be refreshed. Then log corrections, campaign impact and time spent repairing errors; without that record, teams may count the minutes AI saves and miss the cost of correcting its confident mistakes.

Turn AI speed into an accountable advantage

The practical takeaway is not “trust AI” or “ban AI.” Give AI a fast first pass, while requiring evidence and an accountable decision owner wherever the output can change customer experience, spending or brand trust. A small validation step before launch is cheaper than discovering after the campaign that the audience insight was stale, stereotyped or simply unsupported.

Real Internet Sales helps businesses build digital marketing systems that connect AI efficiency with measurable strategy and human judgment. Call 803-708-5514 or visit realinternetsales.com to discuss a practical AI marketing and measurement plan.