If you have a Medicare Advantage plan — and roughly 33 million Americans do as of 2025 — there is a real chance that an algorithm, not a human physician, has already played a role in deciding whether a treatment, hospital stay, or piece of medical equipment gets approved for you. Artificial intelligence has moved from science fiction into the daily machinery of American health care faster than most people realize, and the rules that are supposed to keep it honest are still being drafted, debated, and in many cases, ignored. For Medicare beneficiaries, this is not an abstract technology story. It is a story about who controls access to your care.
The use of AI in Medicare Advantage is most visible — and most controversial — in the prior authorization process. Prior authorization is the requirement that your plan approve certain treatments, procedures, or medications before you receive them. Insurers have long used clinical criteria to make these decisions, but increasingly, they are using AI-driven tools to process and flag requests at scale. A 2023 Senate investigation found that some of the largest Medicare Advantage insurers were using AI tools to deny claims at rates far exceeding what human reviewers would have approved, sometimes overriding the recommendations of the patient's own treating physician. The tool used by one major insurer was reportedly denying post-acute care claims at a rate that internal data showed was clinically inappropriate in a significant percentage of cases.
Dr. Michelle Mello, a professor at both Stanford Law School and Stanford University School of Medicine, and co-director of Stanford's Healthcare Ethical Assessment Lab for AI — known as HEAL-AI — has spent years studying exactly this problem. Her work sits at the intersection of law, ethics, and health policy, and her core message is one that every Medicare beneficiary should hear: the technology is advancing far faster than the accountability structures designed to govern it. Who makes the rules? Who checks whether the AI is getting it right? And critically, who is legally responsible when it gets it wrong? Those questions do not yet have clean answers.
At the federal level, the Centers for Medicare and Medicaid Services — CMS — has taken some steps. In 2024, CMS issued updated Medicare Advantage regulations that explicitly stated that coverage decisions must be made based on the individual patient's circumstances, and that AI or algorithmic tools cannot be used as the sole basis for a denial. The rule was a meaningful signal, but enforcement is another matter. CMS does not have the staff to audit every prior authorization decision made by every Medicare Advantage plan across the country. The agency relies heavily on plan-reported data, beneficiary complaints, and periodic audits — a system that critics argue is too slow and too reactive to catch algorithmic harm in real time.
The Food and Drug Administration has authority over AI tools that qualify as medical devices — things like AI that reads a radiology image or flags a cardiac arrhythmia. The FDA has cleared hundreds of AI-enabled medical devices in recent years, and it has developed frameworks for evaluating them. But the AI tools that insurers use to process prior authorization requests typically do not fall under FDA jurisdiction, because they are classified as administrative rather than clinical tools. That regulatory gap is significant. An AI that tells a radiologist where to look on a scan is regulated. An AI that tells your insurer whether to approve the scan in the first place may not be.
Data Snapshot: According to CMS.gov data, Medicare Advantage enrollment reached approximately 33.8 million beneficiaries in 2025, representing more than 54% of all Medicare-eligible Americans. CMS data also shows that in 2024, there were 3,959 Medicare Advantage plans available nationwide, with an average of 43 plans per county — meaning beneficiaries in most markets have real choices, but also face real complexity in evaluating how each plan uses technology in its coverage decisions. Star ratings, which CMS publishes annually at Medicare.gov, now include measures related to appeals and access to care that can serve as indirect indicators of how aggressively a plan uses denial-based tools.
For beneficiaries trying to protect themselves right now, the most important thing to understand is your right to appeal. If your Medicare Advantage plan denies a prior authorization request or a claim, you have a federally guaranteed right to appeal that decision — and the process has multiple levels. The first level is a redetermination by the plan itself, which must be completed within 60 days for standard requests or 72 hours for expedited requests involving urgent medical need. If the plan upholds the denial, you can escalate to an Independent Review Entity, then to an Administrative Law Judge, then to the Medicare Appeals Council, and ultimately to federal court. The system is cumbersome, but it works — and studies consistently show that beneficiaries who appeal denials win a significant portion of the time. The problem is that most people don't appeal, often because they don't know they can or don't know how.
To file an appeal, start by requesting a written explanation of the denial from your plan — this is called an Explanation of Coverage or a Notice of Denial of Medical Coverage. The notice must explain the specific reason for the denial and include instructions for how to appeal. Your doctor's office can often help you submit a letter of medical necessity, which is one of the most effective tools for overturning an AI-generated denial. If you need help navigating the process, your State Health Insurance Assistance Program — known as SHIP — provides free, unbiased counseling. You can find your local SHIP counselor at shiphelp.org or by calling 1-800-MEDICARE.
Beyond appeals, there are things you can do when choosing or evaluating a Medicare Advantage plan that may help you avoid the worst AI-driven denial practices. CMS publishes Star Ratings for every Medicare Advantage plan at Medicare.gov, updated each fall during the Annual Enrollment Period, which runs October 15 through December 7. Plans rated 4 stars or higher generally perform better on measures related to member experience, access to care, and appeals outcomes. A plan's Star Rating is not a perfect proxy for how it uses AI, but it is one of the few publicly available data points that reflects real beneficiary experiences at scale. During the 2025 plan year, roughly 40% of Medicare Advantage enrollees were in plans rated 4 stars or higher, according to CMS data — which means a substantial share were in plans with lower performance scores.
The legislative landscape is also shifting, though slowly. Several bills have been introduced in Congress that would impose stricter requirements on how AI can be used in insurance coverage decisions, including requirements for human review of AI-generated denials and mandatory disclosure to patients when AI was involved in a coverage decision. As of mid-2026, none of these bills has been signed into law, but the political pressure is building. The Senate Finance Committee has held hearings specifically on AI and Medicare Advantage denials, and CMS has signaled that additional rulemaking is likely. Beneficiaries and their advocates who contact their congressional representatives about this issue are contributing to a policy conversation that directly affects millions of people.
It is also worth understanding that not all AI in health care is adversarial to patients. AI tools are being used to detect cancers earlier, flag dangerous drug interactions, and help primary care physicians identify patients at risk for serious conditions before symptoms appear. Stanford's HEAL-AI lab, where Dr. Mello works, is specifically focused on evaluating AI tools proposed for use at Stanford Health Care facilities — assessing them for accuracy, fairness, and potential for harm before they are deployed. That kind of rigorous, ethics-centered evaluation is exactly what consumer advocates say should be standard practice across the industry, not just at leading academic medical centers. The challenge is that the financial incentives in insurance — where AI that denies more claims can save a plan significant money — are not always aligned with patient welfare.
For Medicare beneficiaries, the practical takeaway from all of this is to be an active, informed participant in your own care. When a treatment is denied, ask why — in writing. When your doctor recommends something and the plan pushes back, ask your doctor to document the medical necessity clearly and specifically. Keep records of every interaction with your plan, including dates, names of representatives, and what was said. If you are approaching the Annual Enrollment Period and considering switching plans, look beyond the premium and the drug formulary — look at the plan's Star Rating, its prior authorization requirements for the services you use most, and its track record on appeals. These details are available at Medicare.gov's Plan Finder tool, and they matter more than most people realize.
The broader question — how high should the guardrails for AI in health care be? — is one that researchers like Dr. Mello are working to answer with evidence and rigor. But while that work continues, Medicare beneficiaries are living inside a system where the technology is already deployed and the rules are still catching up. Knowing your rights, understanding the appeals process, and choosing your plan carefully are not just good habits. Right now, they may be your most reliable protection.