AI in Prior Authorization: Faster Care or Faster Denials?

Lead

Prior authorization has long been one of the most frustrating chokepoints in American health care. A doctor recommends a drug, procedure, or facility stay, but coverage may still depend on an insurer deciding in advance that the care is medically necessary. Now AI is being added to that process. Supporters say it could quickly approve clear-cut requests and reduce waste. Opponents fear it may simply automate a system already known for delays and denials.

Key points

• A cost-control tool with real-world costs: Prior authorization is intended to prevent overuse and unnecessary spending, particularly when lower-cost alternatives exist. In practice, many physicians say it delays care, creates paperwork, and can push patients to abandon recommended treatment.
• Physicians are wary of AI: A 2025 American Medical Association survey found that 61 percent of physicians worry AI will worsen denials of treatments they consider necessary. The AMA is calling for detailed clinical reasoning behind denials and greater transparency around the algorithms insurers use.
• CMS is testing WISeR: The Centers for Medicare and Medicaid Services has launched the Wasteful and Inappropriate Service Reduction Model, or WISeR, in six states through December 2031. The program uses machine learning alongside human clinical review to evaluate selected services that CMS considers vulnerable to overuse, fraud, or abuse, including skin and tissue substitutes, electrical nerve stimulator implants, and knee arthroscopy for knee osteoarthritis.
• The incentive structure is controversial: Vendors hired for the WISeR model can receive a share of what CMS calls “averted expenditures.” Critics argue that this could reward companies for blocking care requests, raising old concerns about profiting from reduced patient access.
• Policy signals are mixed: While CMS expands AI-assisted prior authorization in original Medicare, federal officials are also pressing private insurers and Medicare Advantage plans to reduce and streamline prior authorization. CMS Administrator Mehmet Oz has warned insurers that if they do not ease the burden themselves, the government may regulate them more aggressively.

Why it matters

AI could be useful in prior authorization if it is aimed at making appropriate care easier to approve. It might reduce repetitive documentation, flag incomplete submissions, and speed up requests that clearly meet coverage rules. But prior authorization is not just an administrative workflow; it determines whether patients receive care in time.

Existing evidence explains the skepticism. Federal reports have found that Medicare Advantage plans sometimes denied access to services even when requests appeared to meet coverage rules. Appeals can overturn denials, but that does not erase the practical harm of delay. For patients with limited treatment windows, a later reversal may come too late.

The central question is therefore not whether AI can process paperwork faster. It is whether the system will be transparent, clinically accountable, and designed around patient access rather than denial volume. If vendors are rewarded for “averted” spending without strong safeguards, AI may make an already burdensome process more efficient in the wrong direction.

A better model would use AI as an approval accelerator: helping confirm eligibility, explaining missing evidence, and reserving denials for decisions that include clear clinical justification and meaningful human review. Without those guardrails, the technology risks becoming a black box at the exact point where patients most need accountability.

Source: Ars Technica AI