AI in Prior Authorization: Faster Approvals or More Wrongful Denials?

Prior authorization, the process where health insurers must approve certain treatments before paying for them, is already a major source of care delays. Now AI is being proposed as a way to speed up decisions. But a 2025 American Medical Association survey found that 61 percent of physicians worry AI will make things worse, increasing wrongful denials of treatments doctors consider medically necessary. The debate is now less about whether AI will be used and more about whether it can be trusted to make these decisions fairly.

What happened

Detail	Fact
Survey source	2025 American Medical Association
Respondents	Physicians
Key finding	61% of doctors worry AI will increase denials of necessary treatments
Process in question	Prior authorization (insurer pre-approval for covered treatments)

Prior authorization is the requirement that a health insurer sign off on a treatment, prescription, or procedure before it is carried out. Insurers frame it as a cost control tool, a check against overuse and spending on services where cheaper alternatives exist. Most physicians see it differently. According to the American Medical Association’s 2025 survey, a large majority of doctors report that it causes meaningful care delays, and that some patients abandon recommended treatments entirely while waiting for a decision.

Patients who are denied can appeal, but an appeal adds more time to an already slow process. That friction is the problem AI is now being pointed at.

What AI is supposed to fix

The argument for AI in prior authorization is straightforward. Sorting through patient records, coverage criteria, and clinical guidelines is time-consuming when done by humans. AI can do it faster. For claims that are clearly allowable under existing policy, AI could theoretically issue approvals quickly, reducing the wait time that leads patients to give up on care.

That is the optimistic case. The problem is the pessimistic one has a number attached to it. Sixty-one percent of surveyed physicians believe AI will actually increase the rate of wrongful denials, not reduce them. That is not a fringe concern. It reflects a professional consensus that the technology, as currently applied, is more likely to be used to reject claims faster than to approve them fairly.

Why it matters

For most business owners and operators, this story lands in two places. First, employer-sponsored health plans are a significant cost center, and anything that changes how prior authorization works will affect both premiums and employee access to care. Second, the rollout of AI into high-stakes, regulated decision-making is a preview of disputes that will play out across many industries.

The prior authorization case is a useful test of a broader question: when an AI system makes a consequential decision that harms someone, who is responsible? Insurers deploying AI tools are betting that speed and scale will reduce their administrative costs. Physicians are betting that the same tools will be tuned toward denial, not approval, because that is where the financial incentive sits.

The resistance from doctors is already shaping policy conversations. If 61 percent of physicians are on record with concerns, that is the kind of number that reaches regulators and legislators. AI-driven prior authorization may face mandatory audit requirements or approval rate disclosures before it becomes standard practice.

For businesses thinking about where AI integration creates genuine value versus where it creates liability, this case is worth watching closely. Automating decisions that affect people’s health or finances is categorically different from automating a workflow that routes a support ticket.

Our take

The 61 percent figure is the only concrete number in this story, but it is a significant one. When a supermajority of the professionals closest to a process say a new tool will make it worse, that is worth taking seriously before deployment scales up.

The honest read here is that AI is being introduced into a system where the financial incentives already favor denial. Faster AI does not fix that structural problem. It just accelerates whatever the system was already optimized for. If the underlying criteria are fair, AI speeds up fair decisions. If they are not, AI speeds up unfair ones. The technology is neutral on that question.

Businesses building AI tools for similar high-stakes decisions should note the pattern emerging here: physician opposition organized around a specific, measurable concern (wrongful denial rates) is far more effective than vague unease. If your AI deployment does not have clear success metrics that the affected parties agree on upfront, expect the same kind of resistance to form around it.

We cover the broader shape of these debates in our AI news coverage. The prior authorization fight is one of several where AI’s role in consequential decisions is being contested in public and in policy, not just in research papers.

What to do about it

1. Review your company’s health plan prior authorization policies now, before AI tools are introduced by your insurer without notice.
2. Ask your insurer directly whether AI is used in any part of the denial or approval process, and request documentation of the error rate.
3. If an AI-assisted denial is issued, appeal immediately. Human review is still required at the appeal stage in most jurisdictions.
4. Track how long authorization decisions take as a baseline metric, so you can measure whether any new tools actually speed things up.

The practical takeaway: AI in prior authorization may be coming regardless of physician opinion, so knowing your appeal rights and your insurer’s process is the most useful thing you can do right now.

Source: Ars Technica · AI