Prior authorization is one of the most contested pressure points in American healthcare. Before a patient can receive a physician-recommended treatment, a prescription, or a procedure, the insurer must first approve it. The process was designed to prevent unnecessary spending. In practice, it has become a source of significant friction between patients, doctors, and insurance companies. Now, artificial intelligence is being introduced into that friction point, and the question of whether it will reduce harm or amplify it is genuinely open.
A System Already Under Strain
The scale of the problem is not abstract. A Commonwealth Fund survey cited in the source material finds that roughly one in five American working-age adults with private insurance reported that either they or a family member were denied insurance coverage for physician-recommended care in 2025. Of those who experienced a denial, 41 percent said it delayed their care. More than a quarter reported that their health condition worsened as a result.
In Medicare Advantage, the privately administered alternative to original Medicare that now enrolls roughly 55 percent of Medicare-eligible seniors and disabled people, insurers issue millions of full or partial claim denials annually through prior authorization. A 2022 memorandum from the HHS Office of Inspector General found that more than one in ten denials in Medicare Advantage involved beneficiaries who apparently met coverage rules. Patients can appeal, and the data shows that Medicare Advantage plans overturned 81 percent of denials upon appeal in 2024. But appeals take time, and time is often what patients do not have.
A 2025 American Medical Association survey found that 61 percent of physicians worry that AI will make the denial problem worse, not better. That is not a fringe concern. It reflects a structural tension: the same tool that could accelerate approvals could also be used to accelerate rejections.
What the WISeR Pilot Actually Does
The Trump administration’s response to this situation is a demonstration project called WISeR, which stands for Wasteful and Inappropriate Service Reduction Model. Administered by the Centers for Medicare and Medicaid Services, WISeR uses machine learning combined with human clinical review to evaluate services that CMS believes may be vulnerable to overuse, fraud, and abuse. The targeted procedures include skin and tissue substitutes, electrical nerve stimulator implants, and knee arthroscopy for knee osteoarthritis. The pilot runs through December 2031 across six states.
What makes WISeR structurally significant is that prior authorization has rarely been used in original Medicare before. Medicare Advantage plans have deployed it extensively, but original Medicare has operated differently. WISeR represents a meaningful shift in how the federal government manages care decisions for a large population.
Critics have raised two concerns that go beyond the technology itself. The first is operational: investigations cited by researcher Zena Wolf of the Center for Health and Democracy, referencing reporting from the Washington Post, KFF Health News, and the Seattle Times, suggest that in the early months of the pilot, the model has contributed to care delays and denials across all six states. The second concern is structural. Vendors participating in WISeR earn a share of what CMS calls “averted expenditures,” meaning revenue tied to rejected care requests. That incentive structure has drawn sharp criticism from lawmakers, several of whom have introduced resolutions and amendments to block funding for the model.
Health policy analyst Camm Epstein framed the core tension clearly in a statement to Undark: AI should be used to make appropriate care easier to approve, not necessary care easier to deny.
The Deeper Question AI Cannot Answer Alone
Here is what most coverage of this topic misses. The debate about AI in prior authorization is not primarily a debate about technology. It is a debate about incentives, and AI does not resolve that debate. It accelerates whatever direction the incentive structure is already pointing.
If an insurer or a vendor profits from denials, AI-driven prior authorization will produce faster denials at greater scale. If the system is designed to identify clearly eligible claims and approve them without delay, AI can do that too. The technology is capable of both outcomes. Which one materializes depends on governance, transparency, and accountability, not on the sophistication of the algorithm.
The AMA’s position reflects this logic. The organization advocates requiring insurers to provide detailed clinical reasoning when denying coverage, alongside greater transparency about how AI algorithms reach their conclusions. Without that transparency, patients and physicians have no way to evaluate whether a denial reflects a genuine clinical judgment or an automated pattern optimized for cost reduction.
The Trump administration is itself sending mixed signals. While CMS expands AI-driven prior authorization in original Medicare through WISeR, CMS Administrator Mehmet Oz has simultaneously warned private insurers to reduce their use of prior authorization or face federal regulation. The industry has responded with data showing that prior authorization requests declined by 11 percent between June 2025 and April 2026. Whether denial rates have also declined remains unknown.
In Short
AI can process prior authorization requests faster than any human team. Speed is real. But speed in a flawed system does not fix the system. It scales it. The WISeR pilot is the first major test of AI-driven prior authorization in original Medicare, and its early signals are mixed. The structural question, who benefits when a claim is denied, remains unanswered by the technology. Until incentive design, transparency requirements, and accountability mechanisms catch up with the capabilities of the tools being deployed, AI in prior authorization is as likely to entrench existing problems as to solve them.
Based on reporting from Ars Technica.