Insurers promised to fix prior authorization a year ago. Some now say they won't do all of it
The six-part pledge had January 1, 2026 deadlines: fewer services requiring approval, honoring existing approvals for 90 days after a plan switch, and clearer denial explanations. A year on, some insurers are hedging — but a separate CMS rule on AI denials did take effect.
In 2025, dozens of health insurers signed a six-part pledge to reduce the barriers between patients and care their doctors had already recommended. The commitments were specific and had dates attached. Most of those dates were January 1, 2026.
A year later, the reporting is not flattering. Some insurers now say they will not implement all of the promised initiatives, and patients, advocates and clinicians describe little practical change. What did change came from a different direction: a federal rule about how denials are made, not a voluntary promise about how many there are.
- Reduce the number of services requiring prior authorization
- Honor existing approvals for 90 days when a patient switches plans mid-treatment
- Explain denials clearly, including appeal rights
- Clinician review of all medical-necessity denials — an immediate commitment, not a 2026 one
- From 2026, CMS requires a specific reason for every AI-assisted denial, plus published aggregate approval data
(Sources: KFF Health News, American Medical Association, CMS)
The one that would have mattered most
Of the six commitments, the 90-day continuity provision is the one patients feel. Switching insurance in the middle of a treatment course — because of a job change, an open enrollment, a plan leaving the market — has meant re-justifying from scratch a therapy you are already on and already responding to. Honoring the prior approval for 90 days removes that specific cruelty.
It is also the one where implementation is most uneven, because it requires insurers to accept another company's clinical judgment. A commitment made in a press release runs into a claims system that was not built to do that.
The AI rule is the part with teeth
Separately from the pledge, CMS now requires payers to give a specific reason for every AI-assisted denial and to publish aggregate approval data. The design intent is plain: prevent an algorithm from issuing denials at scale without anyone able to see the pattern.
Several states went further in 2026, passing laws that only a licensed physician or health professional may deny a request on medical-necessity grounds, and that AI cannot be the sole means used to deny, delay or modify care. Where you live now materially changes what an insurer is allowed to automate.
This is the more durable change, because it is a rule rather than a pledge. A voluntary commitment can be quietly downgraded. A published approval rate cannot.
What this means for you, concretely
Nothing here removes prior authorization from your life. What it does is give you specific things to ask for:
- Ask for the denial reason in writing, with the code and the plan language it relies on. You are entitled to it, and vagueness is now harder to defend.
- Ask whether a clinician reviewed it. If the denial was medical-necessity based, a licensed clinician was supposed to have looked at it.
- Ask about the 90-day continuity provision by name if you just switched plans mid-treatment.
- Request expedited review when delay would harm you. It is free to ask and much faster.
- Appeal. A meaningful share of appealed denials are overturned, and the overwhelming majority are never appealed at all.
The full playbook — including how step therapy exceptions work and what a letter of medical necessity needs to contain — is in our guide to prior authorization.