The Centers for Medicare and Medicaid Services is piloting AI-driven prior authorization across six states through December 2031, marking the first time original Medicare will lean on machine learning to screen coverage requests. The program, called WISeR — the Wasteful and Inappropriate Service Reduction Model — targets procedures CMS believes are vulnerable to overuse, including skin and tissue substitutes, electrical nerve stimulator implants, and knee arthroscopy for knee osteoarthritis. It arrives as 61% of physicians surveyed by the American Medical Association in 2025 say they expect AI to make denials worse, not better.
Prior authorization is already the friction point most patients and doctors cite when they talk about American health insurance. A Commonwealth Fund survey found roughly one in five working-age adults with private insurance reported that they or a family member were denied physician-recommended coverage in 2025. Among those hit with a prior authorization denial, 41% said it delayed their care, and more than a quarter said their health problem worsened as a result.
The stakes are highest inside Medicare Advantage, the privately run alternative that now covers about 55% of Medicare-eligible seniors and disabled Americans. A 2022 HHS Office of Inspector General memo found that more than one in ten Medicare Advantage denials involved services that appeared to meet coverage rules. In 2024, plans overturned 81% of denials on appeal — a number that suggests either aggressive initial rejections or a broken first-pass process, depending on who is describing it.
“AI should be used to make appropriate care easier to approve, not necessary care easier to deny.”— Camm Epstein, Health policy analyst
Key facts
- 01CMS is running its WISeR pilot — Wasteful and Inappropriate Service Reduction Model — in six states through December 2031, applying AI to original Medicare prior authorization for the first time.
- 02A 2025 American Medical Association survey found 61% of physicians worry AI will exacerbate denials of treatments they deem necessary.
- 03Medicare Advantage plans overturned 81% of denials on appeal in 2024, and a 2022 HHS OIG memo found more than 1 in 10 denials involved services that met coverage rules.
- 04Industry data shows prior authorization requests declined 11% between June 2025 and April 2026, though the denial rate is not disclosed.
- 05WISeR vendors earn a share of what CMS calls 'averted expenditures,' meaning revenue tied to rejected care requests.
The Biden administration tried to compress the timelines in 2024, requiring insurers on most public-sector plans to decide urgent prior authorization requests within 72 hours and non-urgent requests within seven calendar days. Those rules took effect January 1, 2026. Private insurers separately pledged to standardize electronic prior authorization requests by 2027 and to cut the volume of services subject to prior authorization — including colonoscopies and cataract surgeries — by 2026.
WISeR pushes in a different direction. Instead of pulling prior authorization out of the system, it inserts AI into a corner of Medicare where prior authorization has rarely been used. CMS says the model combines machine learning with human clinical review and will 'ensure timely and appropriate Medicare payment for select items and services.' Vendors carrying out the AI screening earn a share of what CMS calls 'averted expenditures' — revenue tied directly to requests that get rejected.
That payment structure is where critics have focused. Investigations cited by researcher Zena Wolf of the Center for Health & Democracy report that in the pilot's early months, WISeR has already caused delays and denials in each of the six states. Several lawmakers have introduced resolutions and amendments to block funding for the program. CMS Administrator Mehmet Oz has separately warned private insurers to ease the prior authorization burden themselves, telling the National News Desk the government will step in if they do not.
“If you don't do it yourselves, then we're going to do it for you.”— Mehmet Oz, CMS Administrator
Insurers have responded with data suggesting they are complying. Industry figures show prior authorization requests fell 11% between June 2025 and April 2026, though the denial rate over the same window has not been disclosed. In an industry-group survey, every responding health plan agreed with the statement that 'AI or algorithms without clinician or practitioner review are not used to deny prior authorization requests that involve medical necessity or clinical considerations.' Insurers also promised more transparency about the clinical reasoning behind denials.
Physician and Healthcare Huddle founder Jared Dashevsky argues that promise misses what AI in this domain is actually being built to do. He says the technology could 'eliminate barriers, reduce administrative waste, give us more time with patients,' but that the deployed systems are aimed elsewhere. Health policy analyst Camm Epstein made a similar point to Undark, framing the design choice as the entire question: build AI to approve appropriate care faster, or build it to deny necessary care faster.
The AMA is pressing for insurers to disclose the clinical reasoning behind each denial and provide transparency into the AI algorithms themselves. Whether that materializes will depend on whether the industry's voluntary pledges hold, and whether CMS enforces its stated standards on WISeR vendors whose revenue depends on rejections. The 81% appeal-overturn rate in Medicare Advantage is the number that hangs over the whole exercise — it is the benchmark against which any AI-driven system will eventually be judged.
AI is genuinely well-suited to prior authorization. Rules-based coverage decisions, structured medical records, and repetitive claim patterns are exactly the kind of workload where a model can move faster than a call center. The open question is not capability but incentive design: a system paid to approve appropriate care will produce different outputs than a system paid a percentage of averted spending, even if the underlying model weights are identical. WISeR is the first large federal test of which incentive wins, and vendors, insurers, and the AI companies selling into healthcare will all be reading the results through 2031.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
