Patients Three Times More Likely to Trust AI in Clinic Portals Than Public Chatbots

Artificial intelligence offering health advice breeds distrust unless it comes from a trusted source. The Connected Health Consumer report, published in July, finds that patients are three times more likely to trust an AI assistant built into their clinic's secure portal than a public chatbot like the widely available conversational tools.

The report, produced by Salesforce, is the third edition of the Connected Health Consumer study. Researchers asked thousands of patients across several continents how they view artificial intelligence in their interactions with the healthcare system, from booking appointments to sharing medical history. The finding is clear: institutional context shapes trust more than the technology itself.

Where the trust gap comes from

According to the report's authors, the gap between trust in a clinic portal's AI and trust in a public chatbot stems from a sense of institutional accountability. When an AI assistant operates within a system owned by a known healthcare provider, patients assume a specific company, specific doctors and specific data-security procedures stand behind it. A public chatbot, however advanced, offers no such guarantee.

That conclusion echoes growing concern about the quality of medical advice from general-purpose language models. Earlier research, including a study from a University of Oxford team, showed that models which excel at medical knowledge tests can still fail in real conversations with patients who struggle to describe their symptoms precisely.

The problem isn't diagnosis, it's the queue

The report finds that patients' expectations of AI in medicine center not on making diagnoses but on solving logistical problems. Half of respondents hang up after ten minutes on hold with their clinic's phone line, and two-thirds have gone without medication at least once because of delays in issuing an e-prescription. As many as 58 percent of patients admit that complicated digital processes lead them to delay or skip needed medical care altogether.

Against that backdrop, 67 percent of respondents said they would rather get instant help from an AI assistant for simple administrative matters than wait to reach a human agent. The report's authors note that these "boring" tasks, registration, reminders, coordinating data between providers, are the most realistic use case for AI in healthcare today, not replacing doctors in the diagnostic process.

The conditions patients set

Openness to AI doesn't mean unconditional trust. The survey found that 89 percent of patients expect constant access to a live human, and 91 percent want a clear right to reject a recommendation generated by an algorithm. Separate questions probed specific concerns: 36 percent of respondents worry about the accuracy of AI-generated diagnoses, and 30 percent worry about the security of their medical data.

Despite these reservations, acceptance is growing at a pace that would have seemed unrealistic not long ago. In 2024, just 2 percent of American adults turned to AI for health matters; today, 61 percent of people worldwide say they feel comfortable using agentic AI in healthcare. Among people with chronic conditions, 65 percent believe a dedicated AI assistant would make it easier to manage their treatment and medication.

What it means for providers and patients

For hospitals and clinics, the report's findings offer a practical takeaway: investing in AI pays off when the system is part of a provider's own branded infrastructure, not when patients are redirected to an external, general-purpose tool. As many as 90 percent of respondents expect automatic data coordination between hospitals and clinics, putting pressure on system integration rather than simply bolting a chatbot onto a website.

For patients themselves, the survey confirms an intuition many already hold: asking a general chatbot about symptoms feels different from using a tool tied to their clinic account, where their medical history is known and verifiable by a doctor. A growing number of providers worldwide, including in Poland, where 13.2 percent of hospitals now use AI compared with 6.5 percent a year earlier, suggests that this distinction will keep shaping how the technology develops.