Radiology has always been an early adopter of technology, but the last two years have moved faster than most billing teams anticipated. Artificial intelligence is no longer a research topic tucked away in academic imaging departments. It is sitting inside PACS workstations, flagging suspicious nodules, triaging stroke cases, and quantifying plaque burden before a radiologist even opens the study. And now, that technology has its own set of billing codes.
For practices that have spent years perfecting their workflow around standard CPT codes for X-rays, CT, MRI, and ultrasound, the arrival of AI-specific imaging codes changes the equation. It’s not just about adding a new line to the fee schedule.
These codes have their own documentation requirements, coverage rules that apply to them as well as to the payers they are submitted to, and in some instances their own reimbursement rules different from those that radiology billers are used to.
Why AI Imaging Codes Exist in the First Place
The American Medical Association and CMS did not create these codes on a whim. These codes were not developed out of nowhere by the American Medical Association or CMS. AI-powered detection and analysis systems have been found to have clinical benefit, especially in stroke imaging, coronary artery calcium scoring, and pulmonary nodule detection.
If the technology has this kind of impact, there must be a means of tracking its use, and there must be a means for practices to be compensated for its use apart from the underlying imaging study.
That disconnect is what many practices tend to overlook as early on as they get started. An AI analysis code will usually be a supplement to the basic imaging code, rather than being a standalone code. The failure to do so, or to put the two together in the wrong way, is one of the quickest routes to denial, or even an audit red flag for incorrect coding.
The Documentation Gap Nobody Talks About
Radiologists are used to dictating findings. What many are not yet used to is documenting whether an AI tool contributed to the interpretation, what the tool flagged, and how that output was incorporated into the final read. Payers increasingly want to see this distinction in the report itself, not just inferred from the fact that a facility owns AI software.
This creates a real gap between clinical practice and billing readiness. A radiologist can use an FDA-cleared AI tool correctly and still generate a claim that gets denied simply because the documentation does not clearly support the add-on code. Billing teams that are not actively coordinating with radiologists on this language are going to see a rise in denials tied to insufficient specificity, not because the care was wrong, but because the paper trail did not match the code.
Payer Variation Is Already a Headache
Initial claim data is coming out of imaging centers and hospital-based radiology practices, and there’s a pattern to it. Typical reasons for denials of AI-assist imaging codes are related to three areas: pairing the imaging code with the primary code; missing or vague documentation of the AI work; and a failure to have prior authorization when it’s actually required by a payer and goes unspoken.
All these are common issues in medical billing as a whole, but they’re impacting radiology practices more than ever these days due to new codes and workflows not having caught up.
There is a common characteristic of the practices dealing with this well. They’re not using AI imaging billing as a side project that’s added onto workflows. They’re treating AI imaging billing as a separate project, not just a side project that’s tacked onto existing projects. They are not waiting for denials to begin before putting in place charge capture templates, training the coders on the new code sets, and developing payer-specific rule sheets, but well before volume increases.
Where Denials Are Already Showing Up
Initial claim data is coming out of imaging centers and hospital-based radiology practices, and there’s a pattern to it. Typical reasons for denials of AI-assist imaging codes are related to three areas: pairing the imaging code with the primary code; missing or vague documentation of the AI work; and a failure to have prior authorization when it’s actually required by a payer and goes unspoken. All these are common issues in medical billing as a whole, but they’re impacting radiology practices more than ever these days due to new codes and workflows not having caught up.
There is a common characteristic of the practices dealing with this well. They’re not using AI imaging billing as a side project that’s added onto workflows. They’re treating AI imaging billing as a separate project, not just a side project that’s tacked onto existing projects. They are not waiting for denials to begin before putting in place charge capture templates, training the coders on the new code sets, and developing payer-specific rule sheets, but well before volume increases.
What Radiology Billing Readiness Actually Looks Like Right Now
Being ready does not mean having every answer memorized. It means having a process. That includes a clear coding crosswalk for which AI add-on codes pair with which primary CPT codes, a documentation checklist radiologists can follow without slowing down their workflow, and a payer matrix that gets updated as commercial insurers publish or revise their policies.
It also means having someone watching denial trends specifically for these codes, since patterns tend to emerge quickly and can be corrected before they become chronic radiology revenue cycle management issues.
For radiology groups without the internal bandwidth to build and maintain all of that in-house, this is exactly the kind of shift that outsourced billing partners are built to absorb.
The Bottom Line
AI-powered imaging is not something that radiologists need to be thinking about in the future – it’s already creating claims, and those claims are being denied at a significant rate. The risk is not in the technology itself. The danger is that the billing process hasn’t caught up to it.
Those who adopt practices that change how they keep their documentation, know the quirks of payers, and develop a proactive denial tracking process will be the ones capturing the reimbursement this technology is built to unlock, rather than bad debt. Learn more:www.doctormgt.com
