The following is a guest article by Dara St. Louis, Executive Vice President at Reach3 Insights
Overall trust in AI for healthcare has dropped to 44%, down from 52% in 2024, according to new digital health research. But among the 14% of Americans who actually use AI for health and wellness, trust sits at 88%.
That 50-point gap reveals something important: most people are being asked to trust something they can’t see, don’t control, and aren’t sure won’t be used against them.
What’s Driving the Hesitation
When we asked people to explain their hesitation, three concerns dominated. The first was fear of inaccuracy and potential harm, specifically concerns about AI “hallucinations” and misdiagnosis. As one Gen X respondent put it: “AI is often wrong. I wouldn’t trust it with my health.” The second was data privacy: how personal health information is collected, shared, sold, or exposed. And the third was the fear that AI dehumanizes healthcare, stripping away empathy and individualized care. One Gen Z respondent said she didn’t want her “healthcare in the hands of a robot” when she’d “rather be cared for by a human with human thoughts and feelings.”
The Pattern Healthcare Keeps Repeating
Electronic health records were supposed to make care seamless, but often made documentation a burden. Patient portals promised empowerment but delivered jargon-filled lab results with no context. Telehealth worked when it solved an obvious problem during the pandemic, then had to prove its value all over again.
The pattern is consistent: technology succeeds in healthcare when it reduces friction without creating new uncertainty. AI can translate complex medical language, coordinate care across fragmented systems and fill gaps when access is slow or expensive. But in healthcare, people adopt tools they trust to behave predictably, not tools with the most impressive capabilities.
What the Research Shows About Building Trust
Understanding why trust is falling requires a different kind of listening. We used a conversational research methodology for this study, combining quantitative data with open-ended responses and video feedback captured in real time on mobile devices. That approach surfaced the emotional texture behind the numbers: the language in an AI response that felt cold instead of helpful, the confusion about who to call when something seemed wrong. Those details don’t emerge in a standard five-point scale, but they reveal the specific moments where trust breaks down.
The data points to several patterns that separate experiences people trust from those they abandon.
Low-risk entry points drive adoption. Among current users, the most common applications are using chatbots for medical questions or symptom checking (55%), getting personalized health or fitness recommendations (35%), and interpreting test results or lab data (27%). Among people open to using AI, the highest interest is in scheduling appointments and reminders (50%), summarizing complex medical information (49%), and understanding insurance coverage (48%). These are friction points people actively want help with, and they don’t ask someone to gamble their safety on a model’s output.
Visible boundaries build confidence. When we asked what would make people more comfortable, respondents talked about knowing when AI is uncertain, having clear paths to human care when symptoms sound urgent, and understanding what the tool can and can’t do. As one millennial put it: “I would need to know that there are limitations and that it would be very clear when I need to see an actual doctor.” The research suggests that AI that knows when to stop is more reassuring than AI that always has an answer.
Privacy transparency matters more than privacy policies. The data privacy concerns that surfaced in open-ended responses were specific and practical: What data are you collecting? Can I use this without giving you everything? Will this show up in my record? Is it being sold? People want answers upfront, in plain language, with meaningful choices.
Accountability shapes trust. One recurring theme was the fear that no one would be responsible when AI gets something wrong. The responses suggest people are more open to AI when there’s an obvious handoff to a human when it matters and when they know who to reach if something feels off.
Confidence is the metric that matters. Utilization and satisfaction scores can mask underlying trust issues. The gap between users (88% trust) and non-users (38% trust) suggests the critical question is whether people understand what just happened, know what to do next, and feel more in control or more anxious after an interaction.
Trust shifts over time. A one-time study at launch won’t catch the moment when early enthusiasm turns into quiet abandonment or when a small friction point compounds into a trust issue. This is where ongoing research approaches like insight communities become valuable because they allow you to track how sentiment evolves as people have more experiences with AI, as features change, and as news cycles create new concerns. That continuous feedback loop gives you the early signals you need to course-correct before adoption stalls.
The Real Risk
There’s a version of this story where healthcare sprints ahead with AI deployment, hits utilization targets and checks the innovation box, then watches trust erode as people feel surveilled or dismissed. We’ve seen that movie before. It ends with backlash, regulatory intervention, and years spent trying to win back credibility.
The smarter play is to treat trust as infrastructure and build it into workflows, interfaces, data practices, and ongoing research, such as voice-of-market approaches. Trust requires continuous monitoring, not just a launch study. Organizations need to understand how people’s confidence shifts as they use AI over time, where friction emerges, and what reassures them at each stage of adoption, whether that’s in a chatbot, a care journey or an AI-enabled medical device in the home. In the medical devices arena especially, people may be more willing to embrace AI when it clearly improves convenience or monitoring, but less forgiving when setup feels confusing, alerts feel alarming, or the technology creates more stress than support.
Consumers aren’t waiting to be convinced that AI is smart. They’re waiting to be shown that it’s safe, human-centered, honest, and accountable when things go wrong. The organizations that keep listening to the market and to patients in real time (not just at launch) will be the ones that define what responsible AI in healthcare actually looks like.
Dara St. Louis is an Executive Vice President at Reach3 Insights, a full-service consultancy specializing in conversational insights. With over 20 years of experience in market research, Dara is a leader in CPG, tech, retail, health, and experiential insights, known for driving innovation and team empowerment through creative, tech-accelerated solutions in qualitative, quantitative, and community-based research.
No comments:
Post a Comment