Six in ten U.S. consumers say the word "AI" in brand messaging puts them off, according to a WordPress VIP survey published on 17 June 2026. The study polled 1,200 U.S. consumers and adds measurable weight to something many practitioners have been watching anecdotally: conspicuous AI labelling may now be eroding, rather than building, consumer confidence.

The trust numbers are harder to dismiss. Eighty-six percent of respondents said they do not fully trust AI, leaving a narrow 14% who do. That figure matters for any team currently baking "AI-powered" or "AI-driven" into product copy, customer-facing UI, or marketing campaigns — the presumption that the label signals capability and innovation is not matched by what consumers actually feel.

The survey also puts a clock on chatbot interactions. Seventy-four percent of consumers reported experiencing bot fatigue, and the average threshold sits at roughly 40 minutes of engagement before fatigue sets in. For enterprise deployments running AI-assisted support at scale, that 40-minute figure is operationally concrete: it defines a ceiling on unassisted resolution time before user satisfaction begins to degrade, and it gives CX and product teams a defensible benchmark for when to route to a human agent.

The broader context here is worth considering. The current wave of AI product launches has been accompanied by aggressive front-of-pack branding — "AI" appearing in feature names, navigation labels, and press releases at a density that has no real precedent in previous platform shifts. Cloud software vendors rarely stamped "TCP/IP-powered" on their products; mobile app makers didn't lead with "LTE-enabled" as a consumer hook. The AI label has been treated differently, partly because the underlying capability is genuinely novel, and partly because competitive pressure has pushed vendors to signal differentiation fast. This survey suggests that dynamic may have a shorter shelf life than many product teams assumed.

None of this means the technology is losing consumer relevance — it means the marketing vocabulary around it is. There is a meaningful difference between a consumer who distrusts "AI" as a brand signal and one who distrusts the underlying capability when it actually helps them do something. The 40-minute bot fatigue window, for instance, is not evidence that people reject automated assistance outright; it is evidence that poorly calibrated or over-extended automation grinds them down.

For engineering and product leadership, the practical read is fairly direct. Dropping "AI" from consumer-facing copy while retaining the capability underneath is a trivially low-cost experiment relative to the signal this data provides. The companies most likely to act on it quickly are those already separating their internal AI infrastructure roadmap from their external messaging strategy — a distinction that mature platform vendors have learned to make in previous cycles.

The trust deficit is the stickier problem. Eighty-six percent partial or complete distrust is not a number that brand copy changes will close. Closing it requires consistent, observable reliability over time — the same mechanism that built trust in online payments, cloud storage, and navigation apps, all of which faced comparable consumer skepticism in their early mass-market phases. Those shifts took years, not quarters, and they were driven primarily by accumulated experience rather than by messaging.

WordPress VIP, the enterprise arm of Automattic, has a direct commercial interest in how brands manage web content and digital experience, so the survey should be read in that context. The methodology — 1,200 U.S. consumers — is modest for population-level inference, and the framing of questions around "AI in brand messaging" may not capture how attitudes differ across product categories, age cohorts, or interaction modalities. That said, the directional findings align closely with separate industry research that has tracked a growing gap between vendor enthusiasm for AI labelling and consumer receptivity to it.

What this data gives practitioners is not a mandate to hide AI from users, but a concrete prompt to audit where the label appears and what work it is actually doing. In many cases, the answer will be "very little" — and removing it costs nothing.