NPs Benefit Most From AI Scribes. They Also Carry a Billing Risk Physicians Do Not.

When JAMA published the largest multi-site study of ambient AI scribes ever run, most of the coverage focused on the disappointing headline number. Buried in the subgroup analysis was a finding almost nobody wrote about: advanced practice providers were among the clinicians who benefited most.

That is the good news for nurse practitioners, and it is real. The bad news is structural. An NP's note does something a physician's note does not — it decides who gets paid, how much, and under whose NPI. Ambient AI scribes are built to write a clean clinical narrative. They are not built to protect a billing pathway. And in 2026, with a federal audit of exactly this billing pathway underway, that gap has a price.

This is a guide to closing it.

The finding that matters for NPs

On April 1, 2026, JAMA published "Changes in Clinician Time Expenditure and Visit Quantity With Adoption of Artificial Intelligence–Powered Scribes: A Multisite Study" (Rotenstein LS, Holmgren AJ, Thombley R, et al.; doi:10.1001/jama.2026.2253).

The design is strong: 8,581 ambulatory clinicians across Mass General Brigham, Emory Healthcare, UCSF, Yale New Haven Health, and UC Davis — 1,809 adopters against 6,772 non-adopters, tracked from June 2023 through August 2025.

The averages were modest:

• 13 fewer minutes of total EHR time per day (a 3% relative reduction)
• 16 fewer minutes of documentation time per day (about 10%)
• 0.49 additional visits per week

But the authors reported that the most pronounced improvements showed up among primary care physicians, advanced practice providers, female clinicians, and clinicians who used the scribe on at least half their encounters. Heavy users saw roughly twice the EHR time reduction and three times the documentation time reduction — and only about 32% of adopters used it that intensively.

If you are an NP in primary care, you sit in three of those four buckets at once. The tool is likelier to work for you than for the average clinician in that study.

That matters, because the documentation burden is landing hardest on your profession. The Medscape Nurse Practitioner Burnout & Depression Report 2026 found nearly half of NPs burned out, more than one in two reporting burnout, depression, or both, roughly one in six reporting both — and 63% who do not believe relief is coming.

So the case for adopting an AI medical scribe is stronger for NPs than for almost anyone else. The case for configuring it carefully is stronger too.

Why your note carries weight a physician's doesn't

Under 42 CFR 414.56(c), allowed amounts for the services of a nurse practitioner or clinical nurse specialist "may not exceed 85 percent of the physician fee schedule amount." That 15% haircut is the baseline, and it is the reason two special billing pathways exist — pathways that turn on documentation rather than on clinical work.

• Incident-to billing (office and other non-institutional settings) lets a qualifying NP service be billed under the supervising physician's NPI at 100% of the fee schedule.
• Split (or shared) visits (facility settings) pay the practitioner who performed the substantive portion — 100% if that was the physician, 85% if it was you.

In both cases, the difference between 85% and 100% is decided by what the chart says. Not by what happened in the room. By what the note records about what happened in the room.

An ambient scribe listens to the room. It does not know which pathway you are billing.

Pathway one: incident-to, and the audit already running

Incident-to is longstanding Medicare policy (Medicare Benefit Policy Manual, Chapter 15, §60) codified at 42 CFR 410.26. The conditions are strict and they are conjunctive — all of them, every visit:

• The physician personally performed the initial service and established the plan of care. A new patient, or an established patient with a new problem, does not qualify.
• You are following that established plan, not creating one.
• The physician provides direct supervision and remains involved in the ongoing course of treatment.
• The service is furnished in a non-institutional setting.
• The service is of a type commonly furnished in a physician's office, and furnished in accordance with state law — which, as covered below, is not a formality.

What changed for 2026: the current text of 42 CFR 410.26 provides that the supervising physician's presence "may include virtual presence through audio/video real-time communications technology (excluding audio-only) for services without a 010 or 090 global surgery indicator" (as amended, 91 FR 12079, March 12, 2026). CMS finalized virtual direct supervision permanently in the CY2026 Physician Fee Schedule final rule (CMS-1832-F, published November 5, 2025; effective January 1, 2026), after several years of temporary extensions.

Read the exclusions carefully, because they are where practices get into trouble:

• Audio-only does not qualify. A phone call to the collaborating physician is not direct supervision.
• The connection must be real-time and interactive, not a check-in before and a text message after.
• Services with a 010 or 090 global surgery indicator are carved out.

And now the part that should focus your attention. On November 15, 2024, the HHS Office of Inspector General added "Medicare Part B Payments for Incident To Services" (project OAS-25-01-003) to its Work Plan. The Office of Audit Services is examining whether these payments comply with Medicare requirements — specifically whether services were "an integral part of the physician's services" and "furnished under the physician's direct supervision." OIG's stated rationale is that prior work found improving the transparency of incident-to services "is critical to program integrity efforts."

Incident-to is billed under the physician's NPI, which makes it nearly invisible in claims data. That invisibility is precisely what OIG is auditing. The defense is contemporaneous documentation, visit by visit — not a practice-wide policy that NP follow-ups get billed incident-to.

What the AI scribe will not write for you:

• That the physician established the plan of care at a prior encounter, and which encounter that was
• That a supervising physician was on site (or on live audio-video) and immediately available during your visit — with a name
• That the problem addressed was established, not new
• That today's care followed the existing plan rather than changing direction

An ambient scribe transcribes the encounter. None of those four facts is spoken aloud in the encounter. Every one of them has to come from a template, a macro, or a discipline you enforce yourself.

The most common failure mode is not a wrong note. It is a fluent, well-organized, AI-generated note that reads as an independent NP evaluation of a new problem — and was billed incident-to. That note is the auditor's exhibit A.

Pathway two: split (or shared) visits and the attribution problem

In facility settings, 42 CFR 415.140 governs. A split (or shared) visit is an E/M visit performed in part by a physician and in part by a nonphysician practitioner in the same group. The regulation defines the substantive portion as more than half of the total time spent by the physician and NPP performing the visit, or a substantive part of the medical decision making.

The documentation requirements are specific:

• The medical record must identify both practitioners who performed the visit
• The practitioner who performed the substantive portion signs and dates the record and bills the visit
• Modifier FS goes on the facility claim, whichever practitioner bills

Now consider what an ambient scribe produces by default: a single-voice note. One narrative, one provider, one signature block. The two-practitioner structure the regulation requires is precisely the structure the AI flattens.

Two specific risks follow:

• Time is not captured. If you are billing on time, the note needs the total time and enough detail to establish that your share exceeded half. Ambient scribes rarely capture practitioner time by person, and many do not capture time at all unless you dictate it.
• MDM is not attributed. If you are billing on medical decision making, the note has to show which practitioner performed the substantive part of it. A merged narrative shows neither.

If your scribe supports custom templates — and most in this directory do — build a split/shared variant with explicit fields for both practitioners, total time, your time, and an MDM attribution line. Do not rely on editing the default note into shape at the end of a clinic day. That is exactly when it will not happen.

Pathway zero: scope of practice, which the note also has to respect

Incident-to requires that services be furnished in accordance with state law, so your practice authority is a billing input, not just a licensing one.

The AANP maintains the authoritative map. As of its most recently updated policy brief (September 2025), 27 states plus the District of Columbia, Guam, and the Northern Mariana Islands have adopted full practice authority — "authorization of nurse practitioners to evaluate patients, diagnose, order and interpret diagnostic tests and initiate and manage treatments — including prescribing medications — under the exclusive licensure authority of the state board of nursing."

The rest operate under reduced or restricted models, and several of those impose documentation duties the AI has no idea exist:

• A collaborative practice agreement defining your authorized scope, prescribing authority, and consultation expectations
• Chart review requirements — a set number or percentage of your charts reviewed by the collaborating physician on a defined schedule
• Required physician consultation for defined categories of care, which must appear in the note when it happens

These rules change by legislative session, sometimes annually. Verify your own state against the AANP State Practice Environment map rather than any secondary summary, including this one.

A configuration checklist worth thirty minutes

Before your next clinic day, open your scribe's settings and confirm:

• Your credential block is correct. Several scribes default to a physician-style signature. "MD" under an NP's name is a real problem in a real audit.
• You have separate templates per billing scenario — independent NP visit, incident-to, and (if you work in a facility) split/shared. One template cannot serve all three.
• Attestation language is a macro, not a memory. The incident-to elements and the split/shared attribution line should be inserted structurally, with blanks you fill, not typed from scratch at 7pm.
• Time capture is on if you bill on time.
• The note distinguishes new from established problems clearly enough that a reviewer can tell at a glance.
• You read and edit before signing. Your signature makes the note yours regardless of what generated the draft — see what the research actually says about AI scribe accuracy and hallucinations for why that is not a formality.

One more, easy to overlook: a scribe that nudges your documentation toward higher-complexity coding creates a separate exposure. We covered that dynamic, and the payer-side algorithms watching for it, in AI scribes and coding intensity. If you are rolling a tool out across a group rather than solo, the implementation guide covers why adoption intensity decides whether any of this pays off.

Which tools NPs are actually using

Our nurse practitioner category currently lists eight scribes, and they fall into recognizable groups:

Scribe-first, solo-friendly. Freed AI and OrbDoc target individual clinicians with straightforward pricing and fast setup. Heidi Health sits here too, but with far deeper template customization — directly relevant if you need per-pathway templates.

Scribe plus clinical support. Suki AI, Tali AI, and Glass Health pair documentation with clinical decision support, which matters more for NPs practicing at the top of their license in primary care.

Multi-modality. Nabla Copilot — the only scribe in this directory with a published randomized trial behind it — and Twofold Health span general encounters and therapy notes, useful for NPs carrying behavioral health panels. Heidi and Nabla also cover telehealth.

Side-by-side views: AI medical scribes for nurse practitioners, clinical decision support for NPs, and telehealth platforms for NPs. If budget is the binding constraint, start with the best free and low-cost scribes for solo practitioners.

Ask every vendor the same three questions before you buy: Can I build a custom incident-to template? Can the note attribute time and MDM to two named practitioners? What does the signature block say by default? The answers vary far more than vendor marketing pages suggest.

The honest summary

Nurse practitioners have the strongest case of any clinician group for adopting an ambient scribe — the largest study we have puts advanced practice providers among the biggest beneficiaries, and the burnout data says the need is acute.

They also carry a documentation burden that is different in kind. A physician's note has to be accurate. An NP's note has to be accurate and has to establish which of three billing pathways applies, under active federal audit, with a 15% payment swing riding on the answer.

No ambient scribe on the market solves that second problem for you today. What a good one does is hand back the twenty minutes a day you need to solve it yourself — deliberately, in a template, before the visit rather than after it.

Get the setup right once. Then let the tool do what it is genuinely good at.

---

This article is informational only and is not legal, billing, or medical advice. Medicare policy, state scope-of-practice law, and vendor capabilities change frequently — verify current requirements with CMS, your MAC, your state board of nursing, and the vendor before making billing or purchasing decisions.