Part of our AI Regulation and Compliance News series

The Leveraged Years · Clinical Briefing

Medicare WISeR Runs Prior Auth Through AI. How to Win the Appeal.

Medicare WISeR is a new federal model that routes selected traditional-Medicare services through an AI-assisted prior authorization check before they are approved. It went live on January 1, 2026 in six states, and within months clinicians and patients were reporting denials, errors, and delays. For the physician on the other end of that algorithm, the lesson is blunt: when an AI engine denies or delays care, your appeal letter and the clinical note behind it are now the whole ballgame. This briefing shows how to use Claude to draft a clean, defensible appeal fast, the HIPAA-safe rules that keep you compliant while you do it, and the line AI must never cross.

What WISeR is, and where it is live

WISeR stands for Wasteful and Inappropriate Service Reduction. It is a Centers for Medicare and Medicaid Services model that uses artificial intelligence alongside human reviewers to prior-authorize a defined set of traditional-Medicare services. It launched on January 1, 2026 in six states: Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington. Until WISeR, traditional Medicare largely did not impose the kind of upfront prior authorization that Medicare Advantage plans are known for. That is the change clinicians in those states are now feeling.

By mid-2026 the model was under heavy fire. KFF Health News documented confusion, processing errors, and delays for patients and the doctors caring for them. The American Medical Association and the American Hospital Association came out in opposition. On June 16, 2026 a House Appropriations Committee amendment moved to bar the Department of Health and Human Services from funding traditional-Medicare prior-authorization payment models, and separate legislation was introduced to block WISeR outright. The policy fight is real and ongoing. None of that helps the clinician who has a denial sitting in the queue today. For that, you appeal, and you appeal well.

Why AI prior-auth denials hit physicians the hardest

Prior authorization was already one of the most time-consuming administrative tasks in medicine. An AI-screening layer changes the shape of the problem in two ways. First, volume and speed: an automated engine can issue more denials, faster, than a human-only review ever did, which means more appeals land on the practice at once. Second, the burden of proof shifts onto your documentation. When a system flags a service as potentially wasteful or inappropriate, the way you overturn that is with an appeal that states the medical necessity explicitly, cites the relevant clinical criteria, and lays out a clear, documented rationale. A thin note that assumes the reviewer will fill in the clinical logic is exactly the note that loses.

That is the practical pivot. The fight is not really about whether WISeR is good policy. The fight, at your desk, is about whether your appeal makes the case so cleanly that a reviewer cannot reasonably uphold the denial. Writing that letter well, repeatedly, under time pressure, is exactly the kind of structured drafting where a tool like Claude earns its keep.

How clinicians use Claude to draft faster appeals and medical-necessity letters

The core move is the same one we teach for clinical notes: you give Claude de-identified material and a clear instruction, and it returns an organized first draft you read, correct, and own. For an appeal, the material is the denial reason, the relevant clinical facts, and the criteria you are arguing under. Claude is good at this because a strong appeal is largely structural: state what was denied, state the medical necessity, map the patient's facts to the applicable coverage criteria, and ask for a specific outcome.

What Claude does well here:

• Turning your bullet-point clinical rationale into a tight, well-organized medical-necessity argument.
• Structuring a peer-to-peer or written appeal so it directly answers the stated denial reason instead of talking past it.
• Drafting clear language that maps each clinical fact to the coverage criterion it satisfies.
• Producing a consistent, reusable appeal template you can adapt per case, so you are not starting from a blank page every time.
• Tightening tone to be firm, factual, and free of the filler that weakens a clinical argument.

What it must not do is invent clinical facts, guideline citations, or coverage criteria. If Claude produces a guideline reference or a study you cannot independently confirm, you do not use it. The medical-necessity argument has to rest on the real record and real criteria you verify yourself. This is the same review discipline we cover in our briefing on using AI for clinical notes safely, and it is why the note itself matters so much, which we cover in our AI SOAP notes and scribe accuracy review briefings.

A what-to-do-now appeal workflow

Here is a workflow you can run today. Every step assumes the input has already been de-identified, which the next section covers.

One reusable prompt frame ties it together: "Here is a de-identified Medicare prior-authorization denial and the relevant clinical facts. Draft a medical-necessity appeal that restates the denial reason and answers it point by point, mapping each fact to the applicable coverage criterion. Use only the information I provide. Do not invent facts, guideline citations, or criteria. Request a specific outcome." Save it once; it is most of the work.

HIPAA-safe appeals: de-identify before anything leaves the room

Under the HIPAA Privacy Rule's Safe Harbor method, eighteen categories of identifiers must be removed for data to count as de-identified, including names, geographic detail smaller than a state, dates more specific than year tied to an individual, and contact and record numbers. The practical workflow for AI-assisted appeals:

• Strip every identifier before you paste. Replace the patient with "the patient" or "Patient A." Replace specific dates with relative timing where it still makes the clinical point.
• Keep the clinical signal, drop the identity. The diagnosis, the failed treatments, and the functional impact are what the draft needs; the name and the Medicare number are not.
• For any material that genuinely must stay attached to PHI, use only tooling your organization has vetted under a Business Associate Agreement. A consumer chat interface is not that.
• Re-identify on your side, inside your compliant system, after the draft comes back. The AI never needs to know who the patient is.

What AI does not replace

AI drafts the appeal. It does not practice medicine and it does not carry the responsibility. The boundary is bright and worth stating plainly:

• Clinical judgment stays human. Whether the service is medically necessary is your determination, grounded in the patient in front of you. Claude organizes your argument; it does not decide the medicine.
• The physician sign-off is real accountability. When you sign and submit an appeal, you are attesting to its clinical claims. An AI-assisted draft does not dilute that one bit. Read every word first.
• Citations are non-delegable. Any guideline or coverage criterion in the letter must be one you verified. A confident but fabricated reference can sink the appeal and your credibility.
• The patient's care, not the paperwork, is the point. The reason to draft appeals faster is to get patients treated sooner, not to remove the clinician from the decision.

Go deeper: build the full workflow

This briefing is the appeal. The full method, including the notes that survive automated review on the first pass, prompt libraries, and the review protocol that keeps you safe, is the focus of our physician course.

Not sure where to start? The two-minute course finder quiz points you to the right path for your specialty and workflow.

Frequently asked questions

What is Medicare WISeR and where does it apply?

WISeR (Wasteful and Inappropriate Service Reduction) is a Centers for Medicare and Medicaid Services model that uses AI alongside human reviewers to prior-authorize selected traditional-Medicare services. It went live on January 1, 2026 in Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington, and is now facing congressional and medical-society opposition.

How can AI help me appeal an AI prior-authorization denial?

A tool like Claude can turn your de-identified clinical facts into a structured medical-necessity appeal that restates the denial reason and answers it point by point, mapping each fact to the relevant coverage criterion. You provide and verify every fact and citation, then read, sign, and submit. The model drafts; it does not decide the medicine.

Is it HIPAA-compliant to use AI to draft Medicare appeals?

It can be, if you de-identify before any text reaches a public model. Never paste protected health information into a consumer AI tool. Remove all HIPAA identifiers, draft the de-identified appeal, then re-identify inside your compliant system. For material that must stay attached to PHI, use only tools your organization has vetted under a Business Associate Agreement.