The hard part of a split/shared or incident-to visit was never the software. It’s the record. Two people touch the patient, one bills, and the note has to prove who did which part and, for split/shared, how long each of them spent. Get that wrong and it’s a documentation denial or a clawback on audit. An ambient AI scribe helps here by writing down the who, the what, and the how-long as it happens. It does not decide who bills, and it does not attest for you.
That distinction runs through this whole piece. So let’s be clear up front: the scribe drafts a note in English; your clinician reviews it, adds the attestation, and signs. The billing choice belongs to your clinician and biller, not the tool.
Key takeaways
- For CMS, the “substantive portion” of a split/shared visit is more than half the combined time, or a substantive part of the MDM. The billing practitioner has to have done that part.
- Split/shared billing applies in facility settings (hospital inpatient, hospital outpatient, ED). Incident-to lives in the office and has its own separate rules.
- CMS floated a time-only definition years ago, delayed it, and kept the time-or-MDM option through 2026. Check the current Physician Fee Schedule before writing policy.
- The note must name both practitioners and be signed and dated by the billing professional. A scribe can draft all of that; the clinician verifies and attests.
Split/shared and incident-to are two different rules people mix up
They get lumped together because both involve a physician and an advanced practice provider (an NP or PA) on the same patient. The billing logic is not the same, and neither is the setting.
Split (or shared) visits happen in a facility. Think a hospital inpatient floor, a hospital outpatient department, or the emergency department. A physician and an NPP each contribute to one E/M encounter, and one of them bills it. CMS spells this out in its MLN Matters guidance and the Physician Fee Schedule, and the facility-setting limit is explicit: shared services are billed only in a facility setting.
Incident-to is an office thing. It’s how an established patient’s follow-up, performed by an NPP under a physician’s direct supervision, gets billed under the physician’s NPI at the full rate. The conditions live in federal regulation at 42 CFR 410.26: a noninstitutional setting, an established plan of care the physician set, and direct supervision. As of January 1, 2026, that supervision can be met through real-time audio-video presence, not just being down the hall. Audio-only doesn’t count.
Same two people, different building, different rules. A scribe that drafts the note doesn’t change which rule applies. It just makes the note reflect what actually happened.
What the note has to show for a split/shared visit
Start with the substantive portion, because that word is doing all the work. CMS defines it as more than half of the total time spent by the physician or NPP performing the visit, or a substantive part of the medical decision making. The practitioner who did the substantive portion is the one who bills.
Two things follow from that.
If you’re billing on time, the note needs to support more than half of the combined time, and time both people spend jointly with the patient counts once, not twice. So a running record of who was in the room, doing what, for how long, is the documentation you actually need. Not a vague “seen with the resident” line.
If you’re billing on MDM, the note has to show the billing practitioner performed the substantive part of the decision making. That’s harder to capture as a timestamp and easier to capture as content: who ordered what, who weighed the differential, who made the call on the plan.
CMS also closed an old shortcut. You can no longer use history or exam alone to decide who did the substantive portion, because those aren’t in the split/shared code definitions anymore. And critical care split/shared is a special case: it’s time only, never MDM. If your ICU teams bill shared critical care, that’s a rule your compliance staff should flag separately.
One more requirement from the CY 2026 Physician Fee Schedule: the record has to identify both individuals who performed the service, and the billing professional signs and dates it. Two names in the note, one signature that owns it.
Where an ambient scribe genuinely helps, and where it stops
AI Scribe by Patient Square is an ambient AI medical scribe that listens during the visit and hands back a structured SOAP note, ICD-10 suggestions, and a prescription draft, ready to review and sign about two minutes after the visit.
For split/shared documentation, the useful part is the listening. When the physician steps in during an NPP’s encounter, the scribe hears both voices and can draft a note that attributes who covered which parts of the assessment and plan. That attribution, who did what, is exactly the thing thin notes leave out and auditors ask about. The scribe writes it down while it’s fresh, instead of the physician reconstructing it at 8pm.
Here’s the line we won’t cross, and you shouldn’t let any vendor cross it either. The scribe does not:
- decide who bills the visit or whether split/shared applies,
- select the E/M level,
- calculate whether more than half the combined time was met,
- confirm that direct supervision or an established plan of care existed for incident-to,
- write or sign the attestation.
Those are judgment calls and factual assertions about your workflow. The ICD-10 references the scribe surfaces are suggestions for your clinician to confirm, not codes it posts. It never auto-codes and it never selects a level for you. On the split/shared attestation specifically, the scribe drafts the encounter narrative; the clinician adds the line that says “I personally performed the substantive portion” (or your compliance team’s approved wording) and signs. We think that’s the honest boundary. A tool that claims to attest for the clinician is claiming something no software should.
Incident-to: the note is the easy part, the facts are the hard part
Incident-to trips people up because the documentation people worry about isn’t really where the risk sits. The risk sits in the facts around the visit.
For an incident-to claim to hold, the patient has to be established, not new. There has to be a plan of care the physician set, not the NPP starting from scratch. And the supervising physician has to be present under the supervision standard (physically in the office suite, or, from 2026, via real-time audio-video). A perfectly written note doesn’t create any of those facts. It can only reflect them.
So an ambient scribe helps by producing a clean, complete record of the follow-up encounter itself: what the patient reported, what the NPP found, what was done, how it maps to the existing plan. That’s real value, a legible note beats a rushed one every time. What it can’t do is establish that supervision happened or that the plan of care was already in place. Those are compliance facts your team confirms. If the requirements weren’t met, no note fixes it, and you’d bill under the NPP’s own NPI instead.
Run your incident-to setup past your compliance team and your payer. Commercial payers don’t all follow Medicare’s incident-to rules, and some don’t recognize the concept at all.
A quick reference: who owns what
| Task | AI scribe | Your clinician / biller |
|---|---|---|
| Capture who did which part of the visit | Drafts it from the conversation | Verifies it’s accurate |
| Record time each practitioner spent | Notes what it heard | Confirms and totals it |
| Decide substantive portion (time vs MDM) | Does not decide | Decides |
| Select the E/M level | Does not select | Selects |
| Write the attestation statement | Does not write | Writes and signs |
| Confirm incident-to supervision / plan of care | Cannot confirm | Confirms |
| ICD-10 for the encounter | Suggests, never posts | Confirms and codes |
Read that table as the contract. Everything in the left column is drafting. Everything in the right column is judgment and legal attestation, and it stays with people.
The honest verdict
An AI scribe is worth it here for one specific reason: it captures the who-did-what and how-long that split/shared documentation turns on, at the moment it happens, so your clinician isn’t rebuilding the encounter from memory hours later. That’s a real reduction in documentation risk and after-hours charting. It is not a compliance product. It doesn’t attest, it doesn’t bill, and it doesn’t decide whether split/shared or incident-to even applies.
If your practice’s problem is that split/shared notes are thin and inconsistent, a scribe is a strong fix for the note. If your problem is that the underlying facts (supervision, established plans of care, who actually did the substantive portion) are shaky, software won’t save the claim, and you should sort out the workflow first.
When is skipping the scribe the better call? If your NPPs and physicians already document attribution and time cleanly, and the bottleneck is elsewhere in your revenue cycle, the note isn’t your constraint and a scribe won’t move your denial rate. Buy for the problem you actually have.
If you want to see how the draft handles a two-voice encounter and where the attestation line lands, book a demo and bring a real split/shared scenario. And nothing here is legal or billing advice; confirm your split/shared and incident-to policies with your own compliance team and payer, against the current-year Physician Fee Schedule.
For related reading, our note on audio retention covers what happens to the recording (short version: audio is never stored), and the piece on documentation and claim denials covers the narrow set of denials better notes actually touch. Pricing for the US plans is on the pricing page.