What the AMA’s September 2026 CPT Panel Reveals About AI Coding Rules

The American Medical Association’s CPT Editorial Panel met September 17-19, 2026, at
the Hyatt Regency in Minneapolis for the final meeting of the 2028 CPT code cycle. The
agenda ran to roughly 90 tabs covering 49 proposed Category I codes and 40 proposed
Category III codes. Buried in that volume is a small but telling signal for anyone who
codes, audits, or builds software around CPT: for the first time, one of the applications
under review was revised specifically to add validation questions for AI-generated
output.

What happened in Minneapolis

Per agenda documentation analyzed by Madhani Healthcare Consulting, four of the 49
Category I proposals and three of the 40 Category III proposals involved AI or what the
Panel materials describe as “augmentative algorithmic analysis.” That is a modest share
of the total docket, but it marks a shift from prior cycles, when AI-adjacent proposals
were rarer and less formally scrutinized. Public comments on the agenda closed August 10,
and any Category I codes the Panel approves that require pricing will be surveyed for
the American Medical Association’s RVS Update Committee (RUC) meeting in January 2027 —
the next checkpoint before those codes could appear in a future code set.

The pathology and lab application got a rewrite

The most concrete change sits in Tab 94, the Pathology/Laboratory code application.
According to independent health-policy analysis of the pre-meeting agenda,
that application was revised with “extensive revisions… in part to accommodate
AI-related issues,” adding new questions about the software behind a proposed test or
service and how its performance is validated.

Why pathology first

Pathology and lab codes are a natural place for the Panel to start: many algorithmic
diagnostic tools already route through this application, and validation data is often
part of the regulatory record a sponsor has to produce anyway. Extending a similar
software-validation section to general Category I and III applications outside
pathology has not happened yet, which outside observers have flagged as an inconsistency
worth watching as more AI-driven proposals reach the Panel in future cycles.

The AI-only code question the Panel still hasn’t settled

• Four Category I proposals and three Category III proposals in this cycle involved AI
  or algorithmic analysis components.
• Tab 94 (Pathology/Laboratory) was revised to add software-validation questions —
  the first application-level change of its kind.
• Clinically Meaningful Algorithmic Analyses (CMAA), a proposed framework for coding AI
  services that carry no direct physician work component, did not appear on this
  meeting’s agenda.
• Public comment period for the September agenda closed August 10, 2026.
• Approved Category I codes needing pricing move next to the RUC’s January 2027
  meeting.

CMAA has circulated in CPT policy discussion as a way to code AI systems that perform
diagnostic or analytic work without a corresponding physician service to attach a
work relative value unit to — the traditional way CPT valuation works. Its absence from
the September agenda doesn’t mean the idea is dead; the Panel has signaled it expects to
keep exploring the concept, just not on this cycle’s docket.

Why this matters to coders and compliance teams now

None of this changes a coder’s day-to-day work today — nothing from the September
meeting is live in a code set yet, and the earliest any approved codes would surface is
the 2028 cycle. But the direction is worth tracking for three reasons. First, software
validation language in a code application is a preview of what documentation payers may
eventually expect to support a claim built on that code — coders and CDI teams who get
ahead of what “validated AI output” needs to look like in a chart note will have less to
retrofit later. Second, the split between pathology’s new scrutiny and the rest of the
code set suggests uneven rules by specialty for a while, which raises the odds of
inconsistent claims edits and denials in the interim. Third, the CMAA framework, if it
eventually advances, would represent a genuinely new category of CPT code — one without
a human work component — and would change how compliance teams think about medical
necessity documentation for AI-driven services.

What to watch next

The Panel’s Summary of Panel Actions, typically published some weeks after each
meeting, will show which of the AI-related proposals from this docket were approved,
tabled, or rejected. That document, not the pre-meeting agenda, is what determines
whether any of this reaches the 2028 code set — coding teams tracking CPT changes for
compliance purposes should watch for it rather than acting on the agenda alone.

Keeping pace with a code cycle where AI validation criteria are being written in real
time, specialty by specialty, is exactly the kind of moving target that benefits from
automation built to track it. Medikode’s automated
medical coding platform is built to keep coding and compliance workflows current as
CPT and payer rules evolve, so teams aren’t left reconciling code changes by hand after
the fact.