How do patients feel about their doctor using AI scribes?

Many patients leave their healthcare consultations remembering only part of what was discussed. Research shows that patients can forget almost half of medical information told to them shortly after a visit, particularly when conversations involve new diagnoses, medications or complex treatment plans. With the expansion of telehealth and increasing time pressures in clinical practice, ensuring that patients clearly understand their care instructions has become even more challenging.

Artificial intelligence (AI) is increasingly being explored as part of the solution. AI tools can automatically generate consultation summaries that outline diagnoses, treatment plans and follow-up steps. Shared immediately after a visit, these summaries can help patients revisit key advice and share it with family members, carers, specialists, and if needed, with emergency departments.

For clinicians, such tools offer an attractive possibility: reducing cognitive and documentation workload while improving patient understanding. But as AI begins to generate information from consultations that patients receive directly, an important question arises, how much do patients trust these AI tools?

What patients say about AI in healthcare

To better understand patient perspectives, we conducted a survey of 275 Australian healthcare patients examining attitudes toward AI in digital health and AI-generated consultation summaries. The survey asked about general views on AI in healthcare, and presented two undisclosed consultation summaries — one generated by AI and another by a clinician.

The results suggest that Australian patient attitudes to AI in healthcare are cautiously positive but conditional.

Potential benefits were identified, as one patient noted, “AI would make healthcare service more efficient and time-saving for me as a patient. Instead of waiting longer in line or waiting for a doctor's report, I can receive treatment and diagnoses sooner”.

However, this openness was accompanied by clear concerns such as accuracy, reliability and the secure handling of personal health data. More importantly, there was an emphasis that AI tools should support clinicians rather than replace them. As one patient stated, “AI is cool but medical practice cannot 100% depend on AI — AI is the assistance and functional extension of human medical experts”.

Quality of communication shapes patient trust

One of the most striking findings from the survey was how patients evaluated consultation summaries.

While developers often assess AI systems using technical metrics such as accuracy or completeness, patients tend to judge summaries based on communication quality.

Communication quality includes whether the information was easy to understand, whether the tone felt warm, respectful and supportive (as markers of empathy), and whether the summary reflected the conversation they had with their clinician. Even when the medical information was much the same, the way it was worded affected whether people found a summary clear and trustworthy, or distant and impersonal.

As one patient reflected, “This summary feels more patient-centered as it feels like less of a statistic dump and more of an empathetic summary of the consultation with less egregious data”.

For clinicians, this should not be surprising. Consultation summaries are part of the ongoing communication between clinician and patient. When AI systems generate that communication, they effectively become part of the clinical interaction. 

Trust requires demonstrable clinician oversight

Patients in our survey consistently returned to one key point: AI may assist clinicians, but it should not operate without human oversight.

Many patients said they would feel comfortable receiving AI-generated consultation summaries if they knew their clinician had reviewed them before they were shared. As one patient explained, “If AI summaries and tools were required to be checked over by healthcare professionals before patients had access to them, I would be more likely to use them”.

Patients understand that AI can help clinicians with documentation and communication, but they still expect clinicians to remain responsible for the information they receive.

Supporting a multi‑faceted approach to AI governance in healthcare

Across the world, governments and health systems are developing regulatory frameworks for AI in healthcare. These frameworks typically focus on issues such as validation, safety and data protection. The World Health Organization has warned that generative AI tools must be deployed with strong safeguards for transparency and accountability, while Australia’s Artificial Intelligence Ethics Principles similarly emphasise the importance of human oversight.

These initiatives represent an important step toward responsible AI adoption.

However, most current evaluation approaches for healthcare AI focus primarily on technical validation — for example, assessing model accuracy, safety, and performance against benchmark datasets before clinical deployment.

While these measures are essential, our survey findings suggest they may not be sufficient for patient-facing AI tools. Technical validation alone does not address how patients experience AI-generated communication or how these tools influence trust in healthcare interactions.

For patient-facing applications such as consultation summaries, governance frameworks therefore need to incorporate communication quality and demonstrable clinician oversight alongside technical performance.

Key Insights

For healthcare organisations exploring AI-enabled documentation and communication tools, several key insights are highlighted. 

Evaluation frameworks should include patient-centred measures of communication quality, including readability, clarity, and empathy. Implementation should ensure that clinician oversight is explicit and visible (for example, through a visible clinician signoff stamp), so patients understand that clinicians remain accountable for the information they are receiving.

AI tools should also be introduced in ways that strengthen, rather than weaken, the human aspects of care. Used thoughtfully, automated summaries could help patients better understand their care while reducing administrative burden for clinicians.

AI tools will almost certainly play a growing role in healthcare communication. But our survey findings highlight a simple principle: the success of patient-facing AI tools will depend not only on their technical performance, but also on how well it supports the relationships of trust that underpin clinical care.

Dr Wei Zhou is a Research Fellow at the Faculty of Information Technology, Monash University, Australia

Professor Rashina Hoda is Director of HumanAISE lab at the Faculty of Information Technology, Monash University, Australia

Dr Joycelyn Ling is Director, Research Operations and Engagement at Digital Health CRC, Australia

This research was supported by Digital Health CRC Limited (DHCRC). DHCRC is funded under the Australian Commonwealth's Cooperative Research Centres (CRC) Program.