Health insurance companies, which pay for health services of their patrons, can deny coverage of costs if they deem a procedure to be medically unnecessary. This insurer-guided verification of health costs is known as “prior authorization.”
Of roughly 50 million prior authorization requests sent to insurance companies in 2023, roughly 3.2 million were denied. However, when 11.7% of those denials were appealed, 81.7% of appeals resulted in a reimbursement from the insurer, indicating most of the appealed initial denials of provision were faulty.
Under the guise of reducing unnecessary spending, insurance companies make decisions to deny coverage of drugs or durable medical equipment, such as wheelchairs and breathing devices, to curb their costs. However, patients and providers are the ones who pay the price. In 2024, 93% of physicians said prior authorization delayed patient care and led to negative outcomes, while 87% claimed it was a waste of resources.
The usage of artificial intelligence (AI) has begun to pervade insurance companies’ drug authorization processes, further reducing providers’ ability to decide the best course of medical action for patients. Though AI can be trained to assist in diagnosing and creating treatment plans, its potential benefits must be weighed against the reality that AI tools are denying more claims and jeopardizing patient outcomes.
In the 2025 court case “Lokken et al. v. UnitedHealth Group, Inc.” — in which patients who had acute care coverage terminated accused a national health insurance provider, UnitedHealthcare (UHC), of abusing AI to process insurance requests — the court allowed the plaintiffs to continue suing UHC for possibly breaking its own plan promises, but not for violating federal Medicare laws.
Dr. Bruce Scott, the president of the American Medical Association, said these automated systems are denying more claims.
“Emerging evidence shows that insurers use automated decision-making systems to create systematic batch denials with little or no human review, placing barriers between patients and necessary medical care,” Scott said.
The lawsuit examined naviHealth Predict, an AI tool that analyzed cases using algorithms and data, rather than an individualized approach, to decide if a benefit should be provided. In an overwhelming number of cases, insurance was denied.
UnitedHealthcare and other private health care insurance providers are not the only coverage sources utilizing AI to inspect claims. Government-sponsored insurers are currently developing an AI pilot program to accelerate review of Medicare prior authorization requests across Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. The new program, called the Wasteful and Inappropriate Service Reduction (WISeR) Model, is set to start Jan. 1, 2026.
Concerns have arisen surrounding financial incentives for private AI companies contracted by the government to deny more claims. WISeR can easily be corrupted by employees training AI models to be biased towards denying claims, because companies are at less risk of losing money when they don’t have to compensate patients’ medical benefits. Additionally, the contracting process has been murky, with a government spokesperson not identifying the tech companies involved in the pilot.
All of this is not to deny the U.S. health system is wasteful. In fact, overtreatment accounts for 2% to 8.4% of total U.S. health spending. This is a considerable proportion of the total health spending, which constitutes just under 20% of the United States’ gross domestic product. AI usage has its place in health care — confirming diagnoses assessed by doctors, personalizing treatment plans when properly trained and utilizing resources to break costs down effectively. However, extending that role to include the denial of coverage is a slippery slope that could accentuate the negative elements of government health systems, whose grievances mirror those of private insurance.
Reducing the brunt expense of health care is an important priority, but it should be handled with care. The answer to excess spending is not to encroach on the agency of providers, nor to further slow down an already sluggish process to obtain health services. Promoting value-based care and transparency can enhance patient outcomes without relying on algorithms unequipped to handle case-by-case claims. For example, surprise billing has a more pronounced impact on patient experiences and outcomes than wastefulness in overall health spending. Patients receive added health care costs that were not mentioned at clinic visits and end up paying beyond their capacities in many cases.
In a health care system slowly evolving to replace financial incentives with outcome-centric ones, the verification of coverage is a delicate process. It needs to be handled and verified by nonstakeholders — those who do not stand to benefit from a lower quality of patient care.