Healthcare organisations male substantial AI investments to streamline workflows and improve consumer experiences, but confidence in the technology is weakening among both consumers and clinicians. Rapid expansion into daily life has widened the distance between use and trust, especially where AI-generated insights touch critical activities such as education and clinical care. Several recent surveys point to declining optimism, concerns about accuracy and uncertainty over how AI should fit into decision-making. In healthcare, the implications are direct: AI can become a high-value utility only when users understand its role, see its limits and retain confidence that human judgement remains central. Trust now sits alongside performance, safety and governance as a decisive factor in whether AI investment can deliver practical value.
Declining Confidence Across Users and Clinicians
AI is not meeting user expectations at the same pace as its adoption. Recent survey findings show a downturn in optimism and enthusiasm, even as consumers encounter AI tools more frequently and in more settings. Gallup polling shows that even highly engaged AI users feel less positive than they did a year earlier, with the sharpest changes among Generation Z users, aged 14 to 29.
Around half of Generation Z respondents use generative AI daily or weekly. Yet agreement that they feel excited about AI has fallen by 14 percentage points to 22%, while hopefulness has declined by nine points to 18%. Anger about AI technology has risen by nine points to 31% and anxiety remains at 42%. This matters for healthcare because this age group includes future decision-makers, students, early-career care providers, young parents and people who will soon support ageing relatives with complex needs.
Must Read: What Makes Clinical AI Trustworthy
The decline extends beyond one generation. The Ohio State University Wexner Medical Center found that public support for AI fell from 52% in 2024 to 42% in 2026. Pew Research found that users of AI chatbots for health concerns are far more likely to consider them convenient than accurate, at 48% versus 18%.
Accuracy, Skills and Clinical Credibility
Health-related AI faces particular pressure because convenience alone cannot support adoption when users question accuracy. The real-world performance of tools such as ChatGPT and Claude varies, with some outputs creating safety concerns. For healthcare organisations, those concerns complicate efforts to integrate AI into workflows while maintaining confidence from patients and staff. The gap between convenience and perceived accuracy leaves health-related AI poorly aligned with user expectations.
Clinicians express similar unease. An AMA survey found that 63% of physicians remain more concerned than excited about AI in clinical practice. The same findings indicate that most physicians do not expect AI to increase stress or burnout directly. Their concerns focus more on professional capability and the possibility of deskilling across the clinical community.
Skill loss is a major concern. In the AMA survey, 88% of respondents were at least moderately worried about loss of skills among students, colleagues and themselves. Some physicians rated peers who use generative AI as less competent than those relying only on their own skills. Low confidence in the technology can affect professional perceptions as well as tool adoption. If distrust spreads through a clinical practice, AI may become a source of division rather than support, even when intended to improve efficiency or experience.
Transparency, Restraint and Visible Governance
Transparency remains central to user trust, but simple labelling that AI has contributed to a decision does not meet the full requirement. Consumers and clinicians need the ability to understand how an AI tool reached an answer, including which data informed the output, which inferences were made and how confident the model is in its conclusion.
Mayo Clinic findings show how confidence information can change clinician behaviour. With minimal transparency, clinicians overrode 73.9% of AI outputs. With moderate transparency, the override rate fell to 49.3%. With a confidence-calibrated framework, the overall rate dropped to 33.29%. High-confidence predictions were overridden 1.7% of the time, while low-confidence predictions were rejected 99.3% of the time.
Trust also depends on selective deployment and visible accountability. AI earns confidence when it makes a workflow easier, faster or safer, not when it adds complexity. Careful evaluation of the clinical or consumer journey, with input from each person who touches the workflow, helps determine whether AI fits the use case. Governance needs to identify failures quickly, communicate clearly, address problems with urgency and turn lessons into changed practice. Visible accountability can reassure users that clinicians remain in control of diagnosis and treatment pathways, and that AI outputs do not replace human oversight.
Healthcare AI adoption now depends on more than technical performance or investment scale. The widening gap between use and trust creates a practical barrier for organisations trying to embed AI in care delivery, consumer experience and clinical workflows. Improving transparency, deploying tools selectively, preserving clinician control and making governance visible can help align confidence with capability. Without those elements, AI risks becoming an unacceptable source of uncertainty. With them, healthcare organisations have a clearer path to using AI where it can add value while maintaining the trust of patients and professionals.
Source: Digital Health Insights
Image Credit: iStock
