GAO identifies accuracy and reimbursement risks as AI use grows in medical documentation and coding

A GAO report found clinician use of AI documentation and coding tools rose from 21% in 2024 to 28% in 2026, flagging accuracy risks that can trigger overpayment liability.

Categorized in: AI News Healthcare
Published on: Sep 15, 2026
GAO identifies accuracy and reimbursement risks as AI use grows in medical documentation and coding

The U.S. Government Accountability Office examined the growing use of AI for clinical documentation and medical coding in a July 2026 report, flagging accuracy risks and reimbursement compliance as central concerns for healthcare providers. The report arrives as adoption accelerates: an American Medical Association survey cited by GAO found clinician use of AI for these purposes rose from 21% in 2024 to 28% in 2026.

These tools now handle tasks that directly shape claims submitted to Medicare, Medicaid, and commercial payors. AI "scribes" use ambient listening to record patient-clinician conversations and generate draft clinical notes, sometimes without the patient's knowledge. AI coding tools then analyze medical records and suggest billing codes with increasingly limited human review.

The GAO pointed to reduced administrative burden, more detailed documentation, and greater operational efficiency as potential benefits. But the same automation creates compliance exposure that providers cannot afford to ignore.

What the GAO found on accuracy

The report identified a thin evidence base. "There are relatively few independent studies evaluating the accuracy of AI documentation and coding tools," the GAO said. Inaccurate outputs can affect patient care and lead to either over-reimbursement or under-reimbursement.

AI tools may also increase reimbursement by surfacing diagnoses or services that might otherwise go uncaptured. While that additional revenue may be supported by documentation, it raises a practical question: how are providers verifying that AI-generated notes and codes reflect the services actually furnished? The GAO did not answer that question, but the compliance implications are clear.

Where the liability sits

Responsibility stays with the provider regardless of how a note or code was generated. Claims based on AI outputs that do not match the services provided can trigger overpayment liability. Depending on the circumstances and level of knowledge, False Claims Act exposure may follow.

Ambient listening tools also expand the volume of patient information recorded, transmitted, and stored. That creates additional data security risks. Providers must also navigate state-by-state patient notice and consent requirements before recording encounters.

What providers should build now

The GAO's findings point to several controls that belong in any AI-assisted documentation and coding workflow:

  • Clinical judgment must stay human. AI coding suggestions cannot replace independent clinician decisions about diagnosis, severity, level of service, or medical necessity.
  • Review controls are essential. Providers need processes to confirm AI-generated documentation and codes are accurate and supported by the medical record.
  • Payor rules vary. AI workflows should be checked against Medicare Administrative Contractor requirements and payor-specific documentation, coding, coverage, and authentication standards.
  • Privacy and consent require attention. Patient notice and consent rules for recording differ by state, and the increased data volume from ambient tools heightens security obligations.

For medical records clerks and billing staff adapting to these tools, structured training on AI-assisted workflows is becoming a practical necessity. Programs such as AI for Medical Records Clerks and AI for Medical Billers address the documentation, coding, and revenue cycle skills that GAO's findings suggest providers need to build internally.

Why this matters for healthcare professionals

The GAO report confirms that AI documentation and coding tools are not a future concern. Adoption is already at 28% among surveyed clinicians and climbing. For anyone in health information management, medical billing, or compliance, the immediate task is building the verification and oversight routines that keep AI-generated claims defensible. Written policies, training, human validation steps, and privacy safeguards are no longer optional. They are the minimum for responsible use, and the liability for getting it wrong lands squarely on the provider.


Get Daily AI News

Your membership also unlocks:

700+ AI Courses
700+ Certifications
Personalized AI Learning Plan
6500+ AI Tools (no Ads)
Daily AI News by job industry (no Ads)