Radiology AI Reimbursement in 2026: What New Medicare Payment Changes Mean for Imaging Practices

Introduction

Every radiology conference for the past few years has had at least one AI vendor promising that their algorithm will pay for itself. For imaging practices actually trying to build a 2026 budget, the more useful question is narrower and less exciting: does Medicare separately reimburse the AI tool itself, or does the payment still hinge entirely on the underlying imaging study the AI happens to support?

The CY 2026 Medicare Physician Fee Schedule brought real, verifiable changes for radiology, some of which touch AI directly and some of which affect every imaging practice regardless of whether AI is involved. This article separates the two, so you know exactly what changed, what didn't, and where your practice's AI investment actually shows up, or doesn't show up, on a Medicare remittance.

The Big Picture for Radiology Under CY 2026

Before getting into AI specifically, it's worth understanding the broader fee schedule context every imaging practice is now operating under.

Two conversion factors for the first time

For the first time, CY 2026 splits the Medicare conversion factor based on a practitioner's participation in a qualifying Advanced Alternative Payment Model. The conversion factor for qualifying APM participants is $33.5675, reflecting a 0.75 percent annual update. For everyone else, it's $33.4009, reflecting a 0.25 percent annual update. Both figures include a one-time 2.5 percent increase from budget reconciliation legislation and a small budget neutrality adjustment. Since many radiology practices aren't structured to participate in a qualifying APM, most imaging groups will land on the lower, non-APM conversion factor.

Specialty-level impact

CMS's own specialty impact estimates for CY 2026 project an approximate 2 percent decrease for radiology overall, a 1 percent decrease for nuclear medicine, a 1 percent decrease for radiation oncology, and a 2 percent increase for interventional radiology specifically. Diagnostic radiology, in other words, is facing modest downward pressure even before factoring in anything AI-related.

The Efficiency Adjustment and Why It Matters for AI-Enabled Practices

This is the single most consequential structural change in the CY 2026 rule for radiology, and it has a direct, if indirect, connection to AI adoption.

What the efficiency adjustment does

CMS finalized a first-of-its-kind efficiency adjustment, cutting work RVUs and intraservice physician time by 2.5 percent for nearly all non-time-based services on the fee schedule, including procedures, radiology interpretations, and diagnostic tests. The rationale CMS cited is straightforward: as technology and physician experience improve, services tend to become faster to perform, and CMS believes valuations haven't kept pace with those efficiency gains. CMS plans to reapply this adjustment on a three-year cycle going forward.

What's exempt

Time-based codes are excluded from this adjustment, including evaluation and management visits, behavioral health services, services on the Medicare telehealth list, and maternity global codes. New CPT codes effective January 1, 2026 are also excluded from the adjustment for their first year.

The AI connection

There's a pointed irony here worth understanding. CMS's own analysis, drawing on a 2016 Urban Institute pilot comparing fee schedule time estimates to actual electronic health record and observational data, found that imaging and other test interpretations showed some of the largest discrepancies between assumed and actual physician time. AI-assisted workflows, to the extent they speed up interpretation or triage, are part of exactly the efficiency trend CMS is using to justify this ongoing downward adjustment. In other words, the same technology adoption that might make your radiologists faster is also part of the argument CMS is using to justify cutting the work RVU value of the underlying interpretation itself.

Where Radiology AI Actually Gets Reimbursed Today

This is the question most imaging practices actually care about, and the honest answer is that it depends heavily on which category a specific AI tool falls into.

Bundled into the existing interpretation code, with no separate payment

This is where the large majority of radiology AI tools in active clinical use sit right now. Computer-aided detection for mammography, AI-assisted triage flagging for stroke or pulmonary embolism, fracture detection algorithms, and most AI tools that support rather than replace physician interpretation are generally not separately billable. The radiologist bills the standard interpretation code for the underlying study, and the AI tool's contribution isn't separately reimbursed, regardless of how much it improved workflow, sensitivity, or turnaround time. For these tools, the financial case for adoption has to rest on operational value, throughput, reduced miss rates, faster turnaround, rather than a new revenue line.

CPT Category III codes

Category III CPT codes serve as a data-collection pathway for emerging technology, including a number of AI-driven imaging applications. These codes allow a specific service to be tracked and billed, but critically, Category III codes don't carry a guaranteed Medicare payment amount the way Category I codes do. Whether and how much a payer reimburses a Category III code is largely discretionary, varies by Medicare Administrative Contractor and by commercial payer, and can result in denial as often as payment. Practices using an AI tool billed under a Category III code should not assume consistent reimbursement and should verify payer-specific policy before counting on it as a revenue source.

New Category I codes for specific AI-driven imaging analysis

A smaller number of AI-enabled imaging applications have graduated to dedicated Category I CPT codes with an assigned Medicare payment rate. Coronary atherosclerotic plaque assessment, an AI-driven quantitative analysis of coronary CT angiography, is a concrete example reflected in CMS's CY 2026 rulemaking under CPT code 75577. This represents the clearest current model for how a radiology AI application can move from a workflow add-on to a genuinely separately reimbursable service, but it took years of data collection under a Category III precursor code before reaching that point.

New Technology Add-on Payment for inpatient settings

For AI tools used in the inpatient hospital setting, CMS's New Technology Add-on Payment program under the inpatient prospective payment system offers a separate mechanism for temporary additional payment tied to a specific approved technology. This pathway is limited to inpatient hospital use, requires a specific CMS application and approval process, and is time-limited even when granted, so it functions as a bridge rather than a permanent reimbursement solution.

The Supervision Rule Change That Affects AI-Enabled Imaging Workflows

One CY 2026 change worth flagging even though it isn't AI-specific on its face is the permanent adoption of virtual direct supervision for certain diagnostic tests. Physician offices and independent diagnostic testing facilities can now use real-time, interactive audio and video technology to meet direct supervision requirements for services such as Level 2 contrast administration, rather than requiring the supervising physician's physical presence. Audio-only connectivity does not satisfy this requirement; both audio and video are required.

For practices building AI-assisted or remotely supported imaging workflows, particularly those running centralized reading models across multiple sites, this permanent flexibility matters operationally, since it affects staffing models even though it doesn't change how the AI tool itself gets billed.

New MIPS Value Pathways for Radiology

CMS finalized six new MIPS Value Pathways for the CY 2026 performance period, including separate pathways specifically for Diagnostic Radiology and Interventional Radiology. The Diagnostic Radiology pathway includes six quality measures, three Qualified Clinical Data Registry quality measures, eleven improvement activities, and one cost measure. While participation in a specific MVP doesn't directly generate AI-related payment, it does shape how radiology practices report quality performance, and practices already investing in AI-driven quality or workflow tools may find some overlap between what those tools track and what the new radiology-specific MVPs measure.

Common Mistakes Imaging Practices Make With AI Billing

1. Assuming an FDA-cleared AI algorithm is automatically separately reimbursable by Medicare, when FDA clearance and payer coverage are entirely separate determinations.
2. Billing a Category III code for an AI-driven service without confirming the specific payer's current reimbursement policy for that code.
3. Building a return-on-investment case for a new AI tool around anticipated new billing revenue, when most currently available radiology AI tools generate value through efficiency and quality rather than separate payment.
4. Overlooking the efficiency adjustment's cumulative effect on work RVUs for standard interpretation codes when projecting 2026 imaging revenue.
5. Not distinguishing between inpatient New Technology Add-on Payment eligibility and outpatient Medicare Part B billing pathways, which are governed by entirely different rules.
6. Assuming a Category I code exists for a given AI application without checking current CPT and CMS payment files directly.

Actionable Tips for Imaging Practices

1. Before adopting a new AI imaging tool, confirm directly whether it corresponds to an existing bundled code, a Category III code, or a Category I code, rather than assuming based on vendor marketing materials.
2. If a tool is billed under a Category III code, check current reimbursement patterns with your top payers before building revenue projections around it.
3. Model your 2026 imaging revenue with the efficiency adjustment's downward pressure on work RVUs factored in, not just the topline conversion factor increase.
4. Track CMS's ongoing rulemaking for radiology-specific AI codes, since this is an evolving area with new Category I codes potentially emerging in future rule cycles.
5. Confirm your practice's diagnostic testing supervision workflows are updated to take advantage of the permanent virtual direct supervision flexibility where it fits your staffing model.
6. Evaluate AI tools primarily on operational and quality value in the near term, treating any separate reimbursement pathway as a potential future upside rather than a current guarantee.

Expert Recommendations

The practices getting the clearest picture of their 2026 radiology economics are the ones separating two questions that often get blurred together: what does this AI tool do clinically and operationally, and does Medicare actually pay separately for it. Those are genuinely different questions with genuinely different answers right now, and vendor pitches don't always make that distinction clear.

It's also worth watching the coronary plaque assessment code's path closely as a template. It moved from a Category III data-collection code to a Category I code with an assigned payment rate over a period of years, driven by accumulated clinical evidence and utilization data. Other AI-driven imaging applications are likely to follow a similar trajectory, which means the CPT Category III codes your practice is using today for emerging AI tools are worth tracking, not dismissing, since some of them may mature into separately payable services in future rule cycles.

On the efficiency adjustment specifically, this is a policy that will recur every three years going forward, not a one-time cut. Building that expectation into long-term financial planning, rather than treating it as a single adjustment to absorb once, will serve imaging practices better than reacting to each cycle individually.

Frequently Asked Questions

Does Medicare separately reimburse radiology AI tools?

It depends on the specific tool. Most AI-assisted imaging tools currently in use are bundled into the existing interpretation code with no separate payment. A smaller number use CPT Category III codes with no guaranteed reimbursement, and a still smaller number have reached dedicated Category I CPT codes with an assigned Medicare payment rate.

What is the CY 2026 Medicare conversion factor for radiology?

CY 2026 introduced two conversion factors for the first time: $33.5675 for qualifying Advanced Alternative Payment Model participants and $33.4009 for everyone else. Most radiology practices will use the non-APM conversion factor.

What is the efficiency adjustment CMS finalized for CY 2026?

It's a 2.5 percent reduction to work RVUs and intraservice physician time for nearly all non-time-based services, including radiology interpretations, based on CMS's assessment that technology and experience have made many services faster to perform than current valuations reflect. CMS plans to reapply this adjustment every three years.

What is CMS's projected impact on radiology reimbursement for 2026?

CMS's specialty-level impact estimates project an approximate 2 percent decrease for radiology overall, a 1 percent decrease for nuclear medicine, a 1 percent decrease for radiation oncology, and a 2 percent increase for interventional radiology.

Is there a specific CPT code for AI-driven coronary plaque assessment?

Yes. CPT code 75577, covering coronary atherosclerotic plaque assessment, reflects an AI-driven analysis of coronary CT angiography and is included in CMS's CY 2026 Physician Fee Schedule rulemaking as a Category I code with an assigned payment rate.

What's the difference between a CPT Category III code and a Category I code for AI imaging tools?

Category III codes serve as a data-collection pathway for emerging technology and don't carry a guaranteed Medicare payment amount, with reimbursement varying by payer and Medicare Administrative Contractor. Category I codes have an established, assigned payment rate under the fee schedule.

Does FDA clearance of an AI algorithm guarantee Medicare reimbursement?

No. FDA clearance addresses safety and effectiveness for marketing purposes and is entirely separate from a Medicare coverage or payment determination. An FDA-cleared AI tool can still lack any dedicated reimbursement pathway.

What changed with supervision requirements for diagnostic imaging in 2026?

CMS made permanent the ability for physician offices and independent diagnostic testing facilities to meet direct supervision requirements for certain diagnostic tests, such as Level 2 contrast administration, using real-time interactive audio and video technology rather than requiring the physician's physical presence. Audio-only connectivity does not satisfy this requirement.

Are there new MIPS reporting pathways for radiology in 2026?

Yes. CMS finalized six new MIPS Value Pathways for the CY 2026 performance period, including separate pathways for Diagnostic Radiology and Interventional Radiology, each with its own set of quality measures, improvement activities, and a cost measure.

How can an imaging practice get separate reimbursement for a New AI algorithm used in the hospital?

For inpatient hospital use, CMS's New Technology Add-on Payment program under the inpatient prospective payment system offers a mechanism for temporary additional payment tied to a specific approved technology, though it requires a separate CMS application and approval process and is time-limited even when granted.

Conclusion

The honest state of radiology AI reimbursement in 2026 is that most tools still don't generate separate Medicare payment, and the fee schedule changes that did land this year, the efficiency adjustment in particular, apply real downward pressure on the underlying interpretation codes AI tools support rather than creating new revenue around them. A handful of applications, coronary plaque assessment among them, have made the journey from data-collection code to genuinely reimbursable service, and that path is worth watching as a model for where other AI tools may eventually land. Until then, the financial case for most radiology AI adoption rests on efficiency and quality, not a new line item on the remittance advice.