Doximity survey finds 52% of physicians use AI daily, but few have clear rules
Clinical use is rising alongside reported learning benefits. Yet most respondents have encountered errors, and 61% worry about trainees becoming too dependent.
Listen to this story
The audio brief
Story brief
3 key pointsAI use among surveyed physicians is moving deeper into care: 63% reported using it for clinical decision support, up from 37% in 2025, while 70% use it to search medical literature. Yet only 28% said their organizations have clear AI policies, despite widespread permission to use the tools. The gap leaves physicians navigating clinical use amid accuracy concerns and legal uncertainty, making workplace guidance and accountability important issues to watch as adoption expands.
- 01
Doximity surveyed 3,379 U.S. physician members in September 2026; the results describe that sample, not all U.S. physicians.
- 02
Three in four respondents encountered a factual inaccuracy at work in the past year, and two-thirds said they verify clinical output all or almost all of the time.
- 03
Eighty-one percent said AI had taught them something clinically new, but this reflects physicians’ assessments, not measured improvements in performance.
AI is becoming a daily clinical resource before many doctors have clear workplace rules for using it. Doximity’s Doctor-Driven AI Report, released November 12, found that 52% of surveyed physicians use AI daily. Only 28% said their organizations have clear policies and approvals in place, even as use increasingly reaches clinical decisions.
The company surveyed 3,379 of its U.S. physician members in September 2026, with respondents spanning 20 specialties. The findings capture doctors’ reported habits, benefits and concerns—not a clinical trial measuring whether AI improves patient outcomes. They also describe this member sample, rather than establishing how every U.S. physician uses the technology.
From paperwork relief to clinical questions
The initial wave of physician AI adoption focused on administrative work, particularly clinical documentation, according to Healthcare Dive. That benefit remains visible in Doximity’s findings: 46% of respondents said AI had reduced their after-hours administrative work, compared with 36% in 2025. For those physicians, the reported gain extends beyond getting an answer during a patient visit.
Use has also become more frequent. In the latest survey, 85% said they used AI for work at least weekly, versus 51% a year earlier. Daily use rose from 37% to 52%. Oncology led the specialties, with more than 93% of surveyed oncologists reporting at least weekly use.
Clinical uses gain ground
The share of surveyed physicians using AI for clinical decision support rose from 37% in 2025 to 63% in 2026.
Searching medical literature is another major use: 70% said they use AI for that task, up from just over half in the previous year’s survey. Alongside the increase in clinical decision support, that finding places AI in the process of gathering information for care—not only recording what happened afterward.
A learning aid—and a concern for learners
Doctors reported learning from those interactions. Eighty-one percent said AI had taught them something clinically new, including updated guidelines, an alternative diagnosis or a treatment option they had not considered. Separately, 65% said the tools had helped develop their clinical skills and knowledge. These figures reflect physicians’ assessments of what they learned, not independently measured improvements in performance.
That positive view sits beside concerns about dependence. The survey separates worries about doctors’ own skills from those about the people they supervise:
- Trainees: 61% worried that residents, fellows and other trainees would rely too heavily on AI and fail to develop their own clinical skills.
- Their own practice: 30% cited skill erosion as a top limit on AI use. Digital Health Wire reported that the figure reached 43% among the youngest physicians.
Checking answers without clear rules
Accuracy remains a practical problem, not just a hypothetical risk. Three in four respondents said they had encountered a factual inaccuracy while using AI at work during the previous 12 months. More than half were concerned about accuracy and reliability. Two-thirds said they verify AI-generated clinical information all or almost all of the time.
More than a quarter said they spend less effort correcting or double-checking AI output than a year earlier. That reports a decline in checking effort; it does not show whether answers have become more accurate or why those physicians are checking less.
Institutional permission is more common than clear instructions. Doximity said the vast majority of respondents work at organizations that permit or support AI use. Yet many are left to decide how it should fit into patient care. Legal and regulatory uncertainty was the biggest limiting factor for 34% of respondents.
On who should set those boundaries, respondents had a clear preference: 82% said physicians should make the rules for AI in medicine. For now, that preference sits alongside the much smaller share reporting clear policies at work.
Sources
- markets.ft.markitdigital.comMore Than Half of U.S. Physicians Now Use AI In Their Practice Daily, But Few Have Clear Rules For It – Company Announcement
- digitalhealthwire.comDoctor-Driven AI, and AI-Driven Doctors - Digital Health Wire
- healthcaredive.comDespite concerns, AI becomes daily fixture in physician workflows
Reader comments
Newest comments first. Replies stay oldest first.