For years, artificial intelligence lived at the edges of the CPT code set — a handful of Category III codes tracking emerging technology. CPT 2026 marks a turning point: AI-driven services now appear as Category I codes, signaling that algorithm-assisted diagnostics have crossed into standard clinical practice. For coders, that means new codes to learn and a new documentation vocabulary to expect in the record.
Where AI shows up in CPT 2026
The 2026 set validates AI as a routine tool in clinical decision-making, with new codes covering areas such as:
- Coronary plaque assessment. Augmentative software analysis of coronary atherosclerotic plaque from CT imaging.
- Cardiac risk via perivascular fat analysis. Category III codes 0992T and 0993T report AI-based noninvasive cardiac risk assessment using perivascular fat analysis, with 0993T incorporating a concurrent CT scan.
- Burn wound evaluation. Multispectral imaging analysis for assessing burn wounds.
- ECG algorithmic analysis. New codes for algorithm-assisted ECG interpretation — including the 0902T and 0903T–0905T series — plus algorithm-assisted detection of cardiac dysfunction.
Alongside these, CT cerebral perfusion imaging graduated from a long-standing Category III code to new Category I codes — a reminder that today’s emerging-technology codes are tomorrow’s mainstream ones. The broader lab section tells the same story: new Category I analyses like neurofilament light chain testing (83884) and beta-amyloid and tau assays reflect diagnostics that were experimental only a few years ago.
Category I vs. Category III still matters
Don’t let the AI theme blur the fundamentals. Category I codes represent services with established clinical efficacy and typically have assigned RVUs; Category III codes track emerging technology and are frequently non-covered or paid at payer discretion. Before your organization builds volume around any AI service, confirm the code’s category and each major payer’s coverage position. A billable code is not the same thing as a paid code.
The documentation shift
AI-assisted services bring documentation expectations coders haven’t traditionally policed. Increasingly, records supporting these codes should reflect:
- Which algorithm or software was used, and on what data (e.g., the CT dataset analyzed)
- Human oversight — the physician’s review, interpretation, and how the AI output informed clinical decision-making
- The distinct work of the AI analysis versus the underlying imaging or test, so bundled components aren’t double-reported
That last point deserves emphasis. Many AI codes analyze data from a separately reportable study. Read the parentheticals and guidelines carefully to determine when the analysis is an add-on, when it’s bundled, and when reporting both the study and the analysis is appropriate.
What coding teams should do now
- Inventory your organization’s AI tools. Radiology, cardiology, and pathology departments often adopt software before coding hears about it. Find out what’s in use and map it to available codes.
- Verify coverage before go-live. For Category III codes especially, get payer positions in writing and set patient financial-responsibility processes accordingly.
- Educate providers on the oversight narrative. “AI said so” is not documentation. The record should show the clinician’s engagement with the algorithm’s output.
- Track this space annually. AI codes are being added, revised, and promoted between categories faster than almost any other area of CPT.
Medical coding has always translated medicine into data. Now the medicine itself is partly data-driven — and the code set is finally catching up.
CPT is a registered trademark of the American Medical Association. Confirm code descriptors, guidelines, and payer coverage before reporting any service described here.
——————————————————
Originally Published On: Medical Coding News
Photo courtesy of: Getty Images
Follow Medical Coding Pro on Twitter: www.Twitter.com/CodingPro1
Like Us On Facebook: www.Facebook.com/MedicalCodingPro
