Statutory Scope of California AB 3030 in Primary Care
Health and Safety Code §1339.75, enacted by California AB 3030 (Chapter 848, Statutes of 2024) and effective January 1, 2025, applies to any health facility, clinic, physician's office, or group practice office that uses generative artificial intelligence to produce written or verbal patient communications pertaining to patient clinical information. In primary care, that scope reaches portal replies, telehealth chat, result letters, emails, and scripted voicemail advice. Administrative messages about appointment scheduling or billing fall outside the statute's clinical information trigger.
Each qualifying communication must carry two elements. The first is a disclaimer indicating that the content was generated by generative AI. The second is a set of clear instructions describing how the patient may reach a human health care provider, employee, or other appropriate person. The obligation attaches to each communication. A disclaimer on the first message of a thread does not cover later AI-generated replies in the same thread.
Channel-Specific Display Rules
Placement requirements differ by channel. Written physical and digital communications, including letters, emails, and portal messages, must display the disclaimer prominently at the beginning. Continuous online interactions, such as chat-based telehealth, must display it prominently throughout. Audio communications require a verbal disclaimer at both the start and the end. Video communications require a persistent on-screen disclaimer. A notice placed in a footer, a separate consent form, or an EHR metadata field does not meet the prominence standard.
Human Review Exemption and Timestamped Attestation
The statute's exemption is narrow. Section 1339.75 does not apply when a licensed or certified health care provider reads and reviews the AI-generated communication before it is sent. The exemption is only as defensible as the evidence behind it. Legal commentary consistently notes that practices must be able to show who reviewed the message, when, under what credential, and whether the content changed. An unlogged glance at a draft is not an audit trail.
Merry AI records review state at the message level: not reviewed, reviewed and approved, or reviewed and edited. Each state is bound to reviewer identity, license type (MD, DO, NP, PA), and a UTC timestamp captured at attestation and again at transmission. Organizations may then invoke the exemption for attested messages or retain the disclaimer as a conservative default. Many primary care groups keep the disclaimer on behavioral health and heart failure messaging regardless of review status.
Portal Messaging Window Injection
Browser-level injection places the notice in the exact field the patient will read. Merry AI's extension identifies the patient-facing message body, whether a textarea, a contentEditable editor, or an iframe-embedded portal module. It then writes a labeled disclaimer and contact block above any clinical content. It excludes diagnosis, order, and medication fields. Before the message is sent, the extension re-resolves the DOM target after autosave or re-render events, so the notice persists from draft view through send view.
Statutory Comparison: Manual, Generic, and Merry AI
The following comparison isolates liability across the dimensions an auditor or Medical Board investigator would examine.
| Compliance Dimension | Manual Charting | Standard Generic AI Scribes | Merry AI Compliance Architecture |
|---|---|---|---|
| HSC §1339.75(a)(1) disclaimer | Not triggered; no generative AI used | Often absent from patient-facing text | Appended automatically at message start; persistent in chat |
| HSC §1339.75(a)(2) human contact instructions | Written by the clinician as needed | Rarely standardized | Configurable clinic phone, portal, and after-hours routing |
| HSC §1339.75(b) review exemption | Not applicable | Draft edits unlogged | Reviewer, credential, and UTC timestamp logged |
| Field targeting | Manual entry | Note-focused API writes | Patient-facing fields only; structured data excluded |
| Separation from E/M narrative | Clear | Disclosure may mix into note text | Labeled block kept apart from Assessment and Plan |
Generic ambient scribes were designed for visit notes, not outbound patient messaging. When those tools draft patient letters or portal replies, the output frequently reaches the patient without a standardized disclosure. Review status then exists only implicitly in edit history. That gap turns a routine efficiency feature into a statutory exposure on every message.
CMS Documentation Integrity Under AB 3030
Disclosure text must remain distinct from the clinical record that supports reimbursement. Coders and auditors reviewing an encounter should be able to separate regulatory notice language from medical decision making at a glance. Merry AI uses a fixed, labeled header for every AB 3030 block. It never writes disclosure text into Assessment and Plan sections, problem lists, or discrete data elements.
G2211 and Modifier 25 Boundaries
HCPCS add-on code G2211 reflects the longitudinal complexity of serving as the continuing focal point for a patient's care. It is reported with office/outpatient E/M codes 99202-99205 and 99211-99215, and with home or residence codes 99341, 99342, 99344, 99345, and 99347-99350. Since January 1, 2025, CMS permits G2211 payment with a modifier 25 E/M only when the same-day service is an annual wellness visit, vaccine administration, or a Medicare Part B preventive service. AI-drafted portal follow-up that contradicts charted NYHA class, LVEF values, PHQ-9 severity, or Bright Futures periodicity weakens the longitudinal narrative that justifies G2211.
Practices validating this workflow can review peer-reviewed literature on AI-generated patient messaging at https://www.ncbi.nlm.nih.gov/pmc/. They can then map their current portal, letter, and telehealth chat channels against §1339.75 before the next internal audit cycle. To examine your own portal messaging window and attestation trail with our clinical informatics team, Schedule a 15-Minute Workflow Audit.
