Five Major GME Bodies, Zero Guidelines: A New Paper Maps the AI Scribe Governance Gap
A new pilot study out of two Florida family medicine residencies found that ambient AI scribes cut resident documentation time by roughly three minutes per patient and modestly improved measures of resident well-being, according to results published this June in PRiMER. The finding lands alongside a separate paper making a blunter point: none of the field's five major oversight bodies has issued resident-specific guidance on how AI scribes should actually be used in training.
For program directors watching resident burnout metrics, that combination is worth sitting with. AI scribes are moving into residency workflows fast, largely through the same ambient tools already spreading across attending practices, but residents are not attendings. They are still building the clinical reasoning and note-writing judgment that documentation is supposed to teach, and the governance layer meant to protect that hasn't caught up.
The PRiMER study, led by Wayne Anderson, MD, and Jessica Koran-Scholl, PhD, of the University of South Florida and BayCare Health System, followed 13 third-year family medicine residents who used an ambient AI documentation tool over outpatient visits during an intervention period from April to June 2025, part of a broader study window running January through June 2025. Documentation time per patient dropped by about three minutes, a statistically significant result, and residents' scores on the Mini ReZ well-being survey improved as well, particularly on measures of EHR-related stress. Total time spent in the EHR fell initially but rebounded by the study's end, and the authors were candid that a small, uncontrolled pilot cannot settle whether the tool helps residents learn documentation or simply helps them avoid it.
That caveat sits close to the argument made by Julia Giordano and Elizabeth Jones, of the Perelman School of Medicine and Sidney Kimmel Medical College, in a paper published in Advances in Medical Education and Practice this February. They surveyed five major organizations, including the ACGME, AMA, and AAMC, and found none had published guidelines specific to resident use of AI scribes. Their literature review turned up only a small handful of peer-reviewed studies on the subject, of which the new PRiMER pilot is now one. For scale, the authors pointed to Kaiser Permanente's report of more than 15,700 hours of documentation time saved system-wide across 7,260 physicians over 63 weeks, a figure that reflects attending-level adoption rather than residency programs specifically, but signals how fast the underlying tools are spreading regardless of who is using them.
Giordano and Jones's proposed fixes are specific: mandatory resident review and sign-off on every AI-generated note, clear accountability when an error slips through, defined patient consent standards, and limits on which sections of a note AI should draft at all, with assessment and plan left to the resident. None of that exists yet in binding form. The ACGME's most recent Common Program Requirements revision, released this year, does not mention AI.
The tools are already in exam rooms; the guardrails are still theoretical. Whether that gap closes through formal ACGME requirements, specialty society consensus statements, or simply enough pilot data to force the issue is the story worth watching through the rest of this year. What would it take for a program director to trust an AI scribe with a resident's note, and how many more pilots like this one before someone writes that down?
A larger, controlled trial of ambient AI scribes across residency programs, rather than another single-site pilot, is the natural next data point here.
Sourcing
Anderson W, Koran-Scholl J. "The Impact of Ambient AI on Resident Documentation and Well-Being: A Pilot Study." PRiMER, June 18, 2026. journals.stfm.org/primer/2026/anderson-0063
Giordano JA, Jones E. "Resident Use of AI Scribes: A Call for Specialty Society Guidance in Graduate Medical Education." Advances in Medical Education and Practice, Feb 10, 2026. dovepress.com/ai-scribe-use-in-residency-training-a-call-for-specialty-society-guida-peer-reviewed-fulltext-article…
ACGME Common Program Requirements (Residency), 2026 revision (checked directly for AI-related language; none found). acgme.org/globalassets/…/cprresidency_2026_feb_revision.pdf
Kaiser Permanente figure traced to its primary source: Tierney AA, Gayre G, Hoberman B, et al. "Ambient Artificial Intelligence Scribes to Alleviate the Burden of Clinical Documentation." NEJM Catalyst. 2025;6(1). DOI 10.1056/CAT.25.0040 (paywalled; figure independently corroborated via AMA's write-up of the same paper: 15,791 hours saved across 7,260 physicians and 2,576,627 encounters over 63 weeks, Oct 2023-Dec 2024). ama-assn.org/practice-management/digital-health/ai-scribes-save-15000-hours-and-restore-human-side-medicine