CMS MLN905364: Signature Rules for AI Clinical Notes

CMS MLN905364: Signature Attestation Rules for AI-Drafted Clinical Notes

Merry AI · Thoughtfully curated clinical briefs.

CMS MLN905364 (July 2025) confirms that AI scribe output is unsigned until a human clinician attests to it.
The attesting provider authenticates care, not the transcription tool. The scribe need not sign or date.
Merry AI drafts into the EHR field; the clinician applies a valid electronic signature before claim submission.
Group PHP/IOP sessions split cleanly into individually attested progress notes, reducing cloned-note audit exposure.

This brief reads MLN905364 literally. CMS treats artificial intelligence as a scribe. The signature, not the software, carries legal and billing weight. Below is the operational reading for behavioral health and multi-site clinical directors who deploy Merry AI.

• The Loaded Labor & Denominator Model
• Clinical Logic & Audit Defense
• Clinical Taxonomy: ICD-10 Standards
• The Signature Gap MLN905364 Names
• Chrome Extension DOM Overlay
• Clinical Intelligence Layer

The Loaded Labor & Denominator Model

Documentation labor is the hidden denominator in every signature-compliance program. Most groups price scribes against convenience, not fully loaded cost.

A fully loaded medical assistant costs roughly $48,000 per year once a $35,000 base wage absorbs benefits, PTO, and supervision. Merry AI Pro sits at $648 per year.

That ratio lands near 1.3% of a single MA's loaded cost. The attestation workflow stays human; the drafting labor shifts to software.

Review the full cost structure against your panel size on the Merry AI Practice Partner Plans page.

Clinical Logic & Audit Defense

MLN905364 is explicit on authenticity. An unsigned AI draft is not a note. CMS reviewers may disregard it entirely.

Case Study: A multi-site PHP/IOP clinical director audits a 3-hour IOP group with 10 attendees. The shared session audio must become 10 distinct records.

Merry AI Group Note-Splitting converts the shared audio into 10 separate progress notes, each carrying individualized participation, DSM-5-TR observations, risk language, and treatment-plan linkage.

This mirrors the Path Recovery TN deployment, where multi-party group audio resolved into separately attestable records without copy-forward contamination.

Each draft injects into Kipu via the Chrome Extension DOM workflow. The clinician then reviews every note and applies their own valid electronic signature.

This sequence is the audit shield. Signed, human-attested notes prevent CMS reviewers from disregarding AI drafts under MLN905364.

Human-attested clinical metrics carry audits. DSM-5-TR observations, functional scores, and risk assessments attested by a clinician resist California SB 1120 and NCCI Modifier 25 clawbacks.

Individualized attested notes reduce exposure to Joint Commission cloned-note warnings common in copy-forward group documentation. Review peer evidence in JAMA.

Read the primary CMS source directly: CMS Signature Requirements (MLN905364).

For adjacent workflow detail, see our Merry AI Clinical Brief Index.

Clinical Taxonomy: ICD-10 Documentation Standards

Attested notes must map cleanly to coded diagnoses. Behavioral health audits turn on specificity and signature pairing.

Align every coded diagnosis against the CMS National Compliance Standards before signature.

The Signature Gap MLN905364 Names But Does Not Solve

MLN905364 states the rule plainly: sign the entry to authenticate the care, whether a human or AI scribe produced it.

The document stops at principle. It never explains how a closed EHR surfaces an AI draft for that signature in daily workflow.

This is the operational wedge. Compliance depends not on the attestation rule itself, but on where the draft lands before signing.

Our Anchor Truth is simple: documentation integrity holds only when the draft appears inside the EHR field the clinician already signs.

Competitors publish the signature requirement yet skip the injection mechanics. The attestation is only as reliable as the field delivery.

A draft stuck in a separate tab invites copy-paste error and cloned-note flags. A draft placed in-field invites a single, clean signature.

Chrome Extension DOM Overlay & EHR Field Injection

Merry AI runs as a browser-native overlay. There is zero IT setup and no server integration to approve.

The DOM overlay writes drafts directly into existing EHR fields, including closed systems that block traditional API access.

Closed EHR compatibility matters here because the clinician signs inside the native record, satisfying the MLN905364 authentication requirement.

PHP and IOP group note-splitting distributes 10 individualized drafts into 10 field targets, each awaiting its own signature.

Confirm your system coverage within the EHR Clinical Integration Directory.

Clinical Intelligence Layer: Closed-Pilot Orchestration

Signature compliance sits inside a wider workflow. The Clinical Intelligence Layer coordinates pre, during, and post-visit steps.

Pre-visit, the layer assembles prior notes and problem lists so the clinician reviews context before the encounter.

During the visit, ambient capture drafts the note while the provider attends to the patient, not the keyboard.

Post-visit, drafts inject for review and signature, with attestation applied before the claim leaves the building.

The $149 Practice Partner plan runs this orchestration under a closed-pilot structure, selecting five outpatient practices weekly for direct solutions engineering.

Match the plan to your specialty using the Specialty Clinical Playbook Library.

The rule has not changed: the clinician signs. Merry AI simply places the draft where that signature already belongs.