Healthcare
Abridge Details Controls for AI-Generated Physical-Exam Notes

Key PointsAbridge treats clinician review as part of the automation rather than a separate…
View more
- Abridge treats clinician review as part of the automation rather than a separate cleanup step.
- Four controls govern the draft and let clinicians balance templates against spoken findings.
- Assistant prompts adjust a single note while note preferences apply across every visit.
Abridge published clinical guidance on November 19, 2026 explaining how clinicians can combine exam templates, style settings, standing preferences and prompt edits to turn encounter conversations into structured physical-exam documentation.
Casey Hribar, a pediatrician and Abridge Clinical Success Director, wrote the post around an encounter with a two-year-old holding the side of her head. The examination covers both ears, her throat, neck and chest, yet only selected findings are spoken to her parents. The resulting note must distinguish the right ear from the left, include abnormalities supporting the diagnosis and record normal observations from systems actually examined without implying unperformed work.
Those observations help another clinician understand the assessment and give coders, auditors and payers evidence behind coded diagnoses. They also have to fit a structured, terminology-heavy format that differs from the explanations a clinician gives a child and her family.
Templates and Verbal Cues Shape the Draft
Hribar describes conventional electronic health record templates as a partial solution. Starting with a full normal exam meant adding abnormal findings, changing defaults that did not apply and deleting systems not examined. Building from a blank template required entering findings system by system. Both approaches still required checking details against memory of the visit.
Abridge says its system can interpret a remark such as “Okay, now cough… sounds good!” as evidence that a lung examination occurred and, when configured, populate normal findings. By default, verbalized findings replace the corresponding normal language.
Four controls govern the draft. The template determines structure, and the style setting controls how heavily that template is weighed against verbalization. Note preferences add reusable customization, while Assistant prompts adjust an exam after drafting. Clinicians can import an EHR template, build one or choose a specialty-specific option, and switch templates as their work changes.
Specialty templates developed with peer clinicians include obstetrics and gynecology, neurology, mental status, ophthalmology, dermatology and pediatric newborn exams. Clinicians who started using Abridge before a specialty template became available may need to select it themselves.
Clinicians recording through Epic Haiku or Canto can configure the physical exam in the Abridge tab within Hyperspace. Web and desktop users can review the template, style and note preferences through the preferences menu.
Prompt Edits Address Individual Visits
Note preferences operate across visits, whereas Assistant prompt editing applies to a particular note. Telehealth is one case where a clinician’s full default template may not fit. The guidance recommends keeping preference rules specific and checking them on a few notes, and advises against using a preference to insert an entire exam or conditionally apply a saved template.
After a draft is generated, example requests include focusing the exam on an observed skin condition, applying a saved knee-exam template or changing to Limited style. Abridge regenerates the exam and presents proposed changes as tracked edits that clinicians can accept or reject before they are applied to the note.
Frequently used instructions can be bookmarked as saved prompts and reopened through the plus button in the Assistant chat box. That allows a clinician to reuse a request without rewriting it for each note.
The guidance makes clinician review part of the automation rather than a separate cleanup step: the visit conversation supplies findings, while configuration and tracked-edit decisions shape what ultimately becomes the documented exam.












