AI Scribe Coding Concerns 2026: The Upcoding Risk Healthcare Providers Need to Understand

Written by SOAPNoteAI Editorial Team · Updated May 2026

Ambient AI scribes have delivered real benefits: reduced burnout, faster documentation, more time with patients. But a significant and underreported risk has emerged: AI-generated notes are systematically driving higher billing codes — and the gap between documented complexity and actual clinical work is creating serious compliance exposure.

This guide explains what the research shows about AI scribe upcoding, the federal legal risk, and practical steps every practice should take to protect itself.

Create Your Clinical Documentation in 2 Minutes

Start with 20 free SOAP notes. No credit card required.

Try Free on Web

The Problem: AI Documentation Can Outrun Your Clinical Work

AI scribes are designed to capture everything said during a patient encounter and generate comprehensive, complete-sounding clinical notes. This is their strength. It is also the source of the coding problem.

What the Research Shows

A 2025 policy brief published in npj Digital Medicine by researchers at Brown University analyzed coding patterns across multiple health systems before and after ambient AI scribe deployment. The findings were striking:

• Consistent upward shift in E/M coding across all organizations studied
• High-complexity new patient visits (99205) increased by 12–20 percentage points at most organizations
• At one health system, coding for the highest-complexity new patient E/M reached 80% of all new patient visits
• Similar upward trends in established patient visits (99215)

A follow-up analysis by TechTarget's Revenue Cycle Management division found that average claim amounts grew 12–18% after ambient AI scribe adoption, with payers already identifying AI documentation patterns as a fraud indicator.

Why AI Documentation Inflates Complexity

The mechanism is straightforward:

1. AI captures more than you would type. Providers dictating notes manually tend to summarize; AI listening in real time captures every statement the patient makes, every differential diagnosis mentioned aloud, every management option discussed.

2. More documentation = higher MDM. Under the 2021 E/M guidelines, Medical Decision Making complexity is determined by problems addressed, data reviewed, and risk level. A comprehensive AI note often appears to document higher-complexity MDM than a manually typed note of the same encounter.

3. Cognitive default to acceptance. When a note appears complete and well-documented, clinicians reviewing it for signature tend to accept the coded complexity rather than independently assess whether it reflects the actual visit. The note looks like a high-complexity encounter.

4. Commercial incentives. AI vendors are often evaluated partly on documentation completeness and revenue impact. Some tools are explicitly marketed based on revenue lift — creating an incentive structure misaligned with accurate coding.

The Legal Risk: False Claims Act Exposure

Upcoding — billing for a more complex or resource-intensive service than was actually provided — is a federal offense under the False Claims Act (31 U.S.C. §§ 3729–3733).

What "Knowingly" Means

The False Claims Act applies when a provider "knowingly" submits a false claim. Courts have interpreted "knowingly" broadly — it includes:

• Actual knowledge that the claim is false
• Deliberate ignorance of whether the claim is false (e.g., not reviewing AI-generated notes before signing)
• Reckless disregard for the truth (e.g., knowing your AI tool tends to inflate complexity but not auditing)

This means signing AI-generated notes without adequate review is not a safe harbor. The OIG has stated that providers cannot delegate their billing responsibility to AI tools.

Penalties

False Claims Act violations can result in:

• Civil penalties of $13,946–$27,894 per false claim (2026 figures, adjusted annually)
• Treble damages (three times the amount of the false claim)
• Exclusion from Medicare, Medicaid, and other federal programs
• Criminal charges in cases of intentional fraud

For high-volume AI documentation practices, even a modest percentage of upcoded claims can represent substantial liability.

OIG and CMS Guidance

The Office of Inspector General (OIG) and CMS have both emphasized that providers remain responsible for the accuracy of claims regardless of how documentation was created. CMS's 2026 guidance on AI documentation notes that:

"The use of AI tools in clinical documentation does not alter the provider's responsibility to ensure that claims accurately reflect the services rendered and are supported by the medical record."

How to Protect Your Practice

1. Audit Your Coding Distribution Before and After AI Deployment

The single most important protective step is to establish your baseline coding distribution before deploying AI documentation, then audit your distribution 30 and 60 days after deployment.

If your distribution shifts significantly upward — particularly if your 99205/99215 percentages increase substantially — investigate whether your AI tool's documentation is driving inflated complexity.

2. Implement Mandatory Note Review Protocols

Every AI-generated note should receive a focused review of the Assessment and Plan sections before signing, specifically evaluating:

• Number of problems addressed: Does the AI note accurately reflect the conditions actually managed during the visit?
• Data reviewed: Does the AI note accurately capture the external records, tests, or independent interpretation performed?
• Risk level: Is the documented risk level (low/moderate/high) accurate for this encounter?
• Overall MDM level: Does the documented MDM complexity match what you actually did?

3. Train Providers on E/M Documentation Under AI

Most providers have not received training on how AI documentation interacts with E/M coding guidelines. Key principles to reinforce:

• The note supports your coding; your coding reflects your clinical work — not the note's length
• AI-generated comprehensive documentation does not automatically justify a higher-complexity code
• You are responsible for the accuracy of MDM documented in AI-generated notes
• When in doubt, code to what you actually did and adjust the note accordingly

4. Evaluate Your AI Vendor's Documentation Approach

Ask your AI vendor directly:

• Is your tool designed to maximize documentation completeness? If yes, understand how this affects MDM documentation.
• Does your tool suggest coding levels? If so, does it provide conservative suggestions or is it optimized for revenue capture?
• What is your tool's approach to E/M guidelines? Does it apply the 2021 CMS E/M guidelines appropriately?
• Do you provide coding distribution analytics? Good vendors help you monitor for coding drift.

5. Consider Periodic External Audits

For practices with high AI documentation volume or those in higher-risk specialties (primary care, internal medicine, hospitalist medicine), annual external coding audits are prudent. Establish a relationship with a healthcare compliance consultant who understands AI documentation, not just traditional coding patterns.

What Appropriate AI Documentation Looks Like

The goal is not to suppress AI documentation quality — it is to ensure the documentation accurately reflects the clinical encounter. Here is the difference:

Over-Documentation (Red Flag)

A 10-minute established patient visit for hypertension follow-up generates a note with:

• Extensive past medical history review of every problem in the chart
• Review of 12 systems with positive and negative findings for each
• Comprehensive physical exam documentation
• Extensive assessment discussing multiple diagnostic possibilities
• Documentation suggesting high-complexity MDM

If the actual visit was a simple follow-up to renew antihypertensive medication, this note does not accurately represent the encounter — even if everything in it is technically accurate.

Appropriate Documentation

The same visit generates a note with:

• Chief complaint and current medication response
• Relevant ROS for hypertension management
• Pertinent physical exam (vitals, relevant findings)
• Assessment: HTN, well-controlled on current regimen
• Plan: Continue medication, next visit in 3 months
• MDM: Straightforward or low complexity

This note accurately reflects a low-to-moderate complexity visit and supports appropriate coding.

The Path Forward

AI documentation tools offer genuine benefits for provider well-being and care quality. The coding concerns associated with AI scribes are real but manageable with appropriate oversight.

The practices that will navigate this well are those that:

• Maintain human review of AI-generated notes before signing
• Audit their coding distributions proactively
• Train providers on the relationship between documentation and coding
• Choose AI tools that prioritize accuracy over revenue optimization
• Build compliance programs that specifically address AI documentation risks

For documentation tools focused on accuracy rather than billing optimization, see SOAPNoteAI.com — built to help clinicians document accurately and efficiently.