Consumer trust in health care AI depends on more than accuracy
Artificial intelligence in health care can improve workflow, support clinical decisions, and extend access to care, but consumer acceptance cannot be treated as a technical afterthought. In a participatory qualitative study conducted across workshops in Queensland, Australia, 34 consumers described what would make AI acceptable in real care settings and what would make it fail. Their answers point to a central truth: public acceptance of AI in health care is conditional, not automatic, and it depends on whether patients believe the technology serves care rather than replacing it.
Participants described social license as a moving target shaped by three linked forces: relational engagement, structural support, and performance reliability. Those themes were not abstract ideals. They mapped directly onto the way people judge care at the bedside, especially when a tool is introduced into a relationship that already carries vulnerability, uncertainty, and unequal power.
Performance reliability carries the greatest weight in short-term care
For short-term encounters, participants placed the strongest emphasis on whether the system simply works. A tool used to guide triage, flag risk, or assist a time-sensitive clinical decision has little room for error when the patient is in front of the clinician and expecting a clear answer. Performance reliability became the first test of legitimacy: if the system generates poor recommendations, produces opaque outputs, or behaves inconsistently across cases, trust evaporates quickly.
The participants’ reasoning is clinically intuitive. In acute care, patients often measure quality through speed, consistency, and perceived competence. A digital tool that slows the encounter, confuses the care team, or contradicts clinical judgment without explanation threatens confidence in the entire process. Reliability here is not a technical metric alone; it is the patient-facing experience of whether AI can function as a stable clinical aid rather than an unpredictable source of noise.
One participant theme was bluntly practical. If the technology cannot be trusted in urgent care, it has no meaningful place there.
Long-term care requires a different kind of legitimacy
As care becomes longitudinal, the balance shifts. In chronic or ongoing care, relational engagement and structural support become more visible to consumers because patients are no longer judging a single decision. They are judging whether the system respects their history, fits into a sustained relationship with clinicians, and can operate within the broader organization that delivers care.
Relational engagement refers to the expectation that AI should support communication rather than weaken it. Participants wanted tools that preserve time for conversation, empathy, and individualized judgment. In long-term settings, the danger is not only that AI might be inaccurate. It is that the technology could become a barrier between patient and clinician, reducing care to a transaction of outputs and alerts. Structural support mattered as well because consumers understood that even a well-designed tool can fail if staffing, governance, training, or clinical workflow are poor.
This is where the study becomes especially relevant for implementation science. Consumer trust does not depend on a standalone interface. It depends on whether the broader health system has the discipline to regulate the tool, integrate it responsibly, and prevent misuse. People were evaluating the whole environment around the technology, not just the software itself.
Social license is conditional, not a permanent approval stamp
The study frames social license as a dynamic construct rather than a fixed public endorsement. Consumers did not describe a blanket willingness to accept AI in medicine. They described a conditional arrangement in which acceptance depends on evidence that the system will benefit care, respect human relationships, and remain accountable to stakeholders.
That distinction matters for hospitals, health services, and developers. A system that performs well in one context may still fail socially if patients feel excluded from its design or if it is introduced in ways that appear rushed or opaque. Participants favored codesign because they wanted AI built around actual care needs, not grafted onto existing systems as a vendor-led efficiency project. They also wanted stakeholder-driven standards, signaling that legitimacy requires transparent governance and explicit boundaries around use.
The findings suggest a straightforward but demanding standard for implementation: AI should behave as a supportive clinical tool, not a substitute for clinical judgment or a mask for understaffed services. Consumers wanted technology that helps clinicians deliver personalized, empathetic, responsive care. That expectation leaves little tolerance for systems that are presented as neutral while quietly shifting responsibility, reducing human contact, or obscuring decision pathways.
Codesign and governance emerged as the practical path to acceptance
The qualitative data point toward a model in which trust is built before deployment, not repaired after a failure. Codesign gives consumers a role in shaping how AI fits into encounters, how its limits are communicated, and how safeguards are applied. Strict stakeholder-driven standards can define acceptable use, clarify accountability, and reduce the risk that AI is introduced faster than the organization can supervise it.
In clinical terms, this means governance must be visible. Patients need to know when AI is being used, what it is doing, and how clinicians remain responsible for final decisions. They also need assurance that the tool has been aligned with the setting in which it is used. A system designed for one workflow cannot simply be transplanted into another and still expect confidence from consumers.
The study’s design strengthens the relevance of these points. By using participatory workshops and abductive reflexive thematic analysis, the investigators drew out how consumers reason about trust in real-world care rather than in abstract policy language. The sample was modest, but the diversity of ages, countries of birth, disability status, and AI familiarity widened the range of perspectives captured in the discussion.
Consumers judged AI by its effect on care relationships
The central message is practical and clinically consequential: people are willing to consider AI in health care when it reinforces, rather than erodes, the patient-clinician relationship. They want tools that are reliable, governed, and designed with care contexts in mind. They do not appear ready to grant social license simply because a technology is novel or efficient.
For clinicians and health systems, the implication is immediate. AI adoption will be judged not only by performance metrics but also by whether it preserves human connection, protects accountability, and fits the realities of care delivery. If a tool cannot satisfy those expectations, consumer acceptance may remain fragile regardless of its technical sophistication.
The study leaves one operational challenge unresolved: how to translate consumer-defined social license into deployment standards that health systems can actually enforce across different care settings and patient populations.