Tech & Innovation in Healthcare

Focus on the work performed, not the amount of documentation generated.

Healthcare organizations are rapidly adopting ambient artificial intelligence (AI) and digital scribe technologies to reduce documentation burden and allow clinicians to spend more time with patients. These tools listen to patient encounters and generate clinical notes, often producing documentation within minutes of a visit.

Find out how to report codes based on documentation from ambient AI.

Is AI-Generated Documentation Still Provider Documentation?

The potential benefits of AI-generated documentation are significant. Providers may spend less time typing and more time interacting with patients, while organizations may see improvements in efficiency and documentation completion rates. However, the technology also raises important questions for coding professionals, such as:

• Can you code everything documented in an AI-generated note?
• What happens when the documentation includes unsupported diagnoses or conclusions that were never explicitly stated by the provider?
• How should you handle preventive visits that contain extensive discussions of chronic conditions?
• When is a provider query necessary?

While the technology is new, the answers to many of these questions are surprisingly familiar: Coding guidelines, documentation requirements, and compliance standards haven’t changed simply because AI helped create the note.

One of the most important things to remember is that AI-generated documentation should be evaluated using the same standards applied to any other provider documentation.

Whether a note is created through traditional dictation, speech recognition software, a medical scribe, or an ambient AI platform, the provider remains responsible for the content of the medical record. Documentation must be reviewed, authenticated, and maintained according to organizational policies and applicable regulations.

As a coder, your responsibility remains unchanged: Review the record for completeness, consistency, and adequate support for the reported diagnoses and procedures.

Learn What You Can Code

When AI-generated documentation has been reviewed and authenticated by the provider, you may assign codes based on clearly documented and supported conditions, services, and procedures.

Example: If the provider documents “type 2 diabetes mellitus managed with metformin,” “essential hypertension,” and “chronic kidney disease stage 3,” and the record supports the assessment and management of those conditions, you can report the appropriate diagnosis codes just as you would with traditional documentation.

Likewise, procedures, treatments, and services that are clearly documented and supported may be coded according to established guidelines.

The source of the documentation isn’t what matters. What matters is whether the documentation supports the code assignment.

Understand What You Cannot Code

AI-generated notes may sometimes contain information that appears clinically significant but lacks sufficient provider support.

Example: An ambient AI tool might generate statements such as “Possible congestive heart failure,” “Likely pneumonia,” or “Suspected urinary tract infection” without a definitive provider assessment or documented clinical evaluation supporting those diagnoses.

In these situations, you must follow established coding guidelines regarding uncertain diagnoses and outpatient versus inpatient reporting requirements.

Similarly, diagnoses shouldn’t be coded solely because they appear in patient statements.

Example: “The patient states they believe they have long COVID and chronic fatigue syndrome.”

Unless the provider evaluates, assesses, or confirms those conditions, you generally should not code the conditions based solely on the patient’s comments.

You should also be cautious when AI-generated notes include conditions copied from prior encounters. Before assigning codes, verify that the condition was evaluated, monitored, assessed, treated, or otherwise addressed during the current encounter rather than simply carried forward in the documentation.

Spot Unsupported Diagnoses

One of the most important skills in an AI-enabled documentation environment is identifying unsupported diagnoses.

AI systems are designed to summarize conversations and organize information efficiently. However, they may occasionally infer conclusions that extend beyond what the provider actually documented.

Potential warning signs include:

• Diagnoses that appear inconsistent with the clinical documentation may warrant additional review according to organizational policies
• Conditions mentioned only briefly with no further discussion
• Contradictory information elsewhere in the record
• Diagnoses that do not align with laboratory results, imaging findings, or treatment plans
• Copy-forward language that appears unrelated to the current encounter

For example, the AI-generated assessment lists acute kidney injury, but no supporting laboratory findings or discussion appear elsewhere in the record.

Bottom line: When the documentation and clinical picture do not align, additional review is warranted before assigning a code.

Do This When AI Captures a Discussion, Not a Billable Problem

One of the more nuanced challenges associated with ambient AI documentation involves preventive services.

Because AI tools often capture conversations in great detail, they may create documentation that appears to support a separately reportable problem-oriented evaluation and management (E/M) service when the provider simply addressed a routine concern during a preventive visit.

Example: A patient presents for an annual preventive exam. During the visit, the provider notices that the patient’s blood pressure was mildly elevated at check-in and advises the patient to continue monitoring blood pressure at home.

The AI-generated note includes language such as “Hypertension discussed. Patient instructed to continue monitoring blood pressure readings. Follow up if readings remain elevated.”

A coder reviewing the note might wonder whether this discussion supports reporting an office visit E/M service in addition to the preventive service.

Important: The answer depends on the work performed, not the amount of documentation generated.

The mere presence of hypertension-related documentation doesn’t automatically support reporting a separate E/M service. To report a preventive service and a problem-oriented E/M service on the same date, the documentation must support a significant, separately identifiable E/M service beyond the work typically included in the preventive encounter.

For example, support for a separate E/M service might include:

• Evaluation of uncontrolled hypertension
• Review of home blood pressure logs
• Medication adjustments
• Additional diagnostic testing
• Development of a treatment plan
• Significant medical decision making (MDM) related to the condition

Without evidence of separately identifiable problem-oriented work, the presence of additional AI-generated text doesn’t automatically transform a preventive visit into a billable problem visit.

This distinction is particularly important because ambient AI systems may document every topic discussed during an encounter, creating what some organizations refer to as “documentation inflation” or “note bloat.” The focus should remain on the services performed rather than the volume of text captured.

Find out When a Provider Query May Be Appropriate

The need for provider queries doesn’t disappear with AI-generated documentation. In fact, organizations implementing ambient AI solutions may initially experience an increase in queries as providers and coding teams adapt to new documentation workflows.

A provider query may be appropriate when documentation is:

• Ambiguous
• Incomplete
• Conflicting
• Clinically inconsistent
• Missing required specificity

Example: An AI-generated note may indicate that a patient has chronic kidney disease (CKD) but fails to specify the stage. If the required specificity cannot be determined elsewhere in the record, you may need to query the physician.

Similarly, if the documentation references sepsis, respiratory failure, or another significant diagnosis without adequate clinical support, clarification may be necessary.

Recognize Compliance Risks Associated With AI Documentation

While ambient AI tools can improve efficiency, they also introduce several compliance concerns, including:

• Hallucinated content: One of the most widely discussed risks is the potential for AI systems to generate information that was never actually discussed during the encounter. These inaccuracies, often called hallucinations, may sound plausible but lack factual support. If left uncorrected, they can lead to coding errors, reimbursement issues, and audit vulnerabilities.
• Documentation inflation: AI-generated notes may contain more detail than was traditionally documented. While thorough documentation isn’t inherently problematic, excessive detail can create the appearance of work that was not actually performed. This is particularly relevant when evaluating E/M services, preventive visits, and chronic condition management.
• In other words: Increased documentation volume doesn’t necessarily indicate inappropriate billing. However, organizations should ensure that documentation accurately reflects the services performed and MDM involved. You should remain alert for documentation that seems disproportionate to the clinical scenario or that appears to support higher levels of service without corresponding medical necessity.

• Audit exposure: Regulators, payers, and auditors are paying close attention to the use of AI-assisted documentation tools. Records containing unsupported diagnoses, inaccurate summaries, cloned content, or evidence of documentation inflation may face increased scrutiny during audits.

Organizations that fail to establish appropriate oversight processes could face heightened compliance risks, including denied claims, repayment demands, and potential regulatory concerns.

The Bottom Line

Ambient AI and digital scribe technologies are changing how clinical documentation is created, but they aren’t changing the fundamental principles of coding compliance.

As healthcare organizations increasingly integrate AI into clinical workflows, your role becomes even more important. Technology may help create the documentation, but accurate code assignment still depends on careful analysis, sound judgment, and a thorough understanding of coding and compliance requirements.

Suzanne Burmeister, BA, MPhil, Medical Writer and Editor