Summary: AI adoption in healthcare is growing rapidly, but patient trust is not keeping pace. This roundup brings together statistics on physician adoption, AI agents, clinical decision support, documentation, mental health and medication adherence. The findings highlight promising results in specific workflows, alongside the importance of transparency, clinician oversight, and careful evaluation of early evidence.
Table of Contents
Introduction
AI went from pilot project to everyday tool in healthcare faster than almost anyone predicted. Four in five U.S. physicians now use it. Half of healthcare organizations have generative AI in production. And AI agents are already completing prior authorizations without a human touching them.
Patients, meanwhile, are hesitating. Fewer of them are open to AI in their own care than a year ago, and most want to be told when it’s being used.
That gap between fast adoption and slow trust is the real story behind the numbers. Below are the AI in healthcare statistics that matter most, with a focus on conversational AI and AI agents: the tools that talk to patients and clinicians. Every number comes from a peer-reviewed study or a major national survey, and links to its source.
AI in healthcare: the top 10 stats at a glance
- 81% of U.S. physicians now use AI in their practice, up from 38% in 2023. (AMA, 2026)
-
50% of healthcare leaders surveyed said their organization had implemented generative AI. (McKinsey, 2026)
- 61% of health system and health plan tech leaders are already building agentic AI or have budget set aside for it. (Deloitte, 2026)
- 40% of prior authorizations at MUSC Health are completed by AI agents with no human involvement. (Deloitte, 2026)
- 51.9% → 38.8%: clinician burnout dropped after 30 days with an ambient AI scribe, across six U.S. health systems. (JAMA Network Open, 2025)
- 90%: Google’s conversational AI, AMIE, included the final diagnosis among its top seven possibilities in a supervised study of 100 primary care patients. (Google Research, 2026)
- 5,636 patients, 6 trials: AI chatbots produced a medium-sized improvement in medication adherence over standard care. (PLOS Digital Health, 2026)
- 22% of Americans get health information from AI chatbots at least sometimes. (Pew Research Center, 2026)
- 52% → 42%: the share of U.S. adults open to AI in their own care fell between 2024 and 2025. (Ohio State Wexner Medical Center poll, via CIDRAP)
-
About 3 in 4 U.S. adults say it’s extremely or very important that providers tell them when AI is used in their care. (Pew Research Center, 2026)
AI adoption in healthcare: where things stand
AI is becoming part of routine healthcare work, particularly research summarization and clinical documentation. Agentic AI—which can carry out tasks across a workflow—is also moving beyond pilots, although it remains less widely deployed than generative AI overall.
These surveys offer a snapshot of adoption, rather than a census of the healthcare sector:
- 81% of U.S. physicians surveyed use AI in their practice, up from 38% in 2023. Research summarization and clinical documentation are the most common uses. (AMA, 2026)
- 50% of healthcare leaders surveyed reported that their organization had implemented generative AI, up from 25% in late 2023. McKinsey surveyed 150 leaders from payers, care providers, and health technology firms in September–October 2025. (McKinsey, 2026)
- 19% of healthcare leaders surveyed by McKinsey reported implementing agentic AI, while another 51% reported running proofs of concept. (McKinsey, 2026)
- 85% of U.S. physicians surveyed want to be consulted or directly involved in their organization’s AI adoption decisions. (AMA, 2026)
Why it matters: Adoption is advancing, but implementation alone does not demonstrate better care or a return on investment. Providers need to identify useful workflows, measure results and involve the clinicians who will use the tools.
The patient front door: AI agents for intake, triage and patient chat
Conversational AI can help collect symptoms, answer initial questions and prepare patients for a clinical visit. Early studies show encouraging patient experiences in supervised services, but results from structured workflows should not be generalized to patients using a chatbot on their own.
- 81% of responding patients chose Alan’s AI assistant: In French health insurer Alan’s medical chat service, a randomized experiment involving 926 cases found higher satisfaction with AI-assisted chats than physician-only chats, with similar levels of trust. The AI was clearly labeled and physicians supervised the conversations. (Alan, 2025)
- 100 patients completed supervised AI history-taking: At Beth Israel Deaconess Medical Center, Google’s AMIE collected pre-visit histories with live physician oversight. No safety stops were required, and clinicians reported that the summaries helped them prepare for visits. This was an early feasibility study. (Google Research, 2026)
- Participants using AI chatbots identified the relevant condition in fewer than 34.5% of medical scenarios: An Oxford-led study of 1,298 UK adults found that chatbot-assisted participants performed no better than those using their usual resources, despite strong performance by the AI models when tested alone. (Nature Medicine, 2026)
Why it matters: Strong model performance does not automatically translate into useful patient guidance. For intake and patient chat, these findings favor carefully designed services with a clear purpose, transparent AI labeling and a route to clinical support.
Ambient AI scribes and clinician burnout
Ambient AI scribes listen to consultations and draft clinical notes for clinicians to review. Studies suggest they can reduce documentation burden and improve clinician well-being, although time savings vary across users and settings.
- Clinician burnout fell by 13.1 percentage points after ambient AI scribe use: In a study involving 263 clinicians across six U.S. health systems, estimated burnout prevalence decreased from 51.9% to 38.8% after 30 days. The before-and-after design does not establish causation. (JAMA Network Open, 2025)
- Ambient AI scribes reduced documentation time by a median of 2.6 minutes per appointment: A study of 45 clinicians across 17 specialties also reported a 29.3% reduction in after-hours electronic health record work. (Study summarized in npj Digital Medicine, 2025)
- One study found that AI scribes saved an average of 34 seconds per note: Time savings varied substantially between physicians. (Study summarized in npj Digital Medicine, 2025)
Why it matters: The potential benefit extends beyond faster note-taking to less work after hours and more attention during consultations. Providers should measure those outcomes in their own workflows, while retaining clinician review to catch omissions and inaccurate details.
AI for clinical decision support
Conversational AI is showing promise in diagnostic and treatment reasoning, but accuracy is only part of the picture. A useful recommendation must also be practical, appropriate for the patient and integrated into a clinician’s decision-making process.
- Google’s AMIE listed the diagnosis subsequently established by clinicians among its top seven possibilities in 90% of cases: In a supervised study of 100 primary care patients, the conversational AI system’s first-choice diagnostic accuracy was 56%. (Google Research, study, 2026)
- Physicians produced more practical and cost-effective care plans than Google’s AMIE: Blinded evaluators otherwise rated the AI system and physicians similarly on overall diagnostic and management-plan quality. AMIE had no access to patient records or a physical examination. (Google Research, 2026)
- AI chatbot assistance improved physicians’ treatment reasoning: In a Stanford-led trial of 92 physicians, doctors using a chatbot outperformed those using conventional resources alone on clinical management tasks. The trial assessed patient case scenarios, not outcomes in actual care. (Stanford Medicine, 2025)
Why it matters: These findings support further evaluation of AI as a clinical assistance tool. They do not establish equivalent performance across everyday medical care, where patient history, examination findings and clinical judgment remain essential.
Conversational AI in mental health
Clinical trials suggest that chatbots can offer modest improvements in depression and anxiety symptoms, but newer generative AI tools have not yet shown a clear advantage over scripted chatbots.
- AI chatbots reduced depression and anxiety symptoms across 18 trials involving 3,477 adults: A meta-analysis of randomized controlled trials found small but statistically significant improvements in both outcomes, with the clearest benefits around eight weeks. (Journal of Affective Disorders, 2024)
-
Generative AI chatbots showed no clear advantage over scripted chatbots for depression and anxiety: A 2026 meta-analysis found small benefits for chatbots overall, but results for the generative AI subgroup did not reach statistical significance. (npj Digital Medicine, 2026)
Why it matters: Chatbots may offer check-ins and structured exercises between visits. The evidence does not justify treating them as a replacement for professional mental health care, and services need a clear route to human support when distress arises.
AI chatbots and medication adherence
AI chatbots show promise for improving medication adherence, particularly when they proactively engage patients and combine reminders with education and support. A review of randomized trials suggests that the frequency and substance of the conversation matter. Simply making a chatbot available on demand may offer less benefit than regular outreach.
- AI chatbots improved medication adherence across six trials involving 5,636 patients: A meta-analysis of randomized trials found a medium-sized, statistically significant improvement compared with standard care among patients with chronic conditions. (PLOS Digital Health, 2026)
- Regular AI chatbot outreach showed significant medication adherence benefits: Daily or weekly-to-monthly chatbot engagement was associated with improved adherence. On-demand or occasional engagement did not show a significant benefit. (PLOS Digital Health, 2026)
- AI chatbots combining reminders with education and support improved medication adherence: Interventions incorporating education, coaching, and questions and answers alongside reminders showed significant benefits. Reminder-only chatbot interventions did not. (PLOS Digital Health, 2026)
Why it matters: The findings favor services that help patients understand and follow their treatment, rather than simply prompting them to take medication. These subgroup results are useful design signals, although they do not establish which individual feature caused the improvement.
Patient trust and transparency
Patients are turning to AI for health information, but willingness to accept it in their own care has declined. Most want providers to disclose AI use, while many remain unsure whether they have already encountered it.
The surveys reveal a gap between using a chatbot independently and feeling informed about AI within a healthcare service.
- 22% of U.S. adults get health information from AI chatbots at least sometimes, compared with 85% who get it from healthcare providers. Pew surveyed 5,111 adults in October 2025. (Pew Research Center, 2026)
- Openness to AI in healthcare fell from 52% to 42% among U.S. adults between 2024 and 2025. The Ohio State Wexner Medical Center poll also found declining confidence that AI could make healthcare processes more efficient. (Ohio State Wexner Medical Center, via CIDRAP)
- About three-quarters of U.S. adults consider disclosure of AI use in their care extremely or very important. The finding comes from a Pew survey of 3,488 adults conducted in June 2026. (Pew Research Center, via Healthcare Dive, 2026)
- Patients want AI disclosure for routine tasks too: 72% of U.S. adults want to be told when AI takes clinical notes, 64% when it orders prescription refills and 56% when it schedules appointments. (Pew Research Center, via Healthcare Dive, 2026)
- 46% of U.S. adults are unsure whether AI has been used in their care, and 53% feel they have little or no say over its use. These findings highlight gaps in both patient awareness and perceived control. (Pew Research Center, via Healthcare Dive, 2026)
Why it matters: Transparency needs to explain what AI does, how it affects care and where human responsibility sits. Clear labeling, understandable explanations and opportunities to ask questions can help patients make informed choices.
What the numbers add up to
- Adoption and trust are moving at different speeds. AI use is growing among clinicians and healthcare organizations, while patients want clearer explanations and more say in how it affects their care.
- Workflow design matters alongside model performance. Promising results come from specific applications—such as supervised intake and clinical documentation—not simply from giving people access to a capable chatbot.
- Patient engagement needs more than reminders. Medication adherence findings suggest value in regular outreach that combines practical prompts with education and opportunities to ask questions.
- The evidence is promising, but still developing. Short studies, small samples and simulated tasks remain common. Providers should distinguish early findings from demonstrated improvements in everyday care.
For healthcare teams, the practical next step is to choose a clearly defined workflow, measure its results and make the role of AI understandable to the people using it.
Looking to bring conversational AI into your healthcare workflows? QuickBlox combines chat, video consultations and configurable AI agents to support patient intake, answer routine questions and connect patients with your team. Get in touch to discuss your use case and explore how QuickBlox could fit your service.
Talk to a sales expert
Learn more about our products and get your questions answered.
Contact sales Additional resources
Explore these guides for a closer look at how healthcare AI works in practice:
AI Agent — automate conversations and workflows. Now available on Free plan
