guide
AI prior authorization: where it works for providers and payors
AI prior authorization is the use of artificial intelligence on either side of a prior authorization request. Practices use it to read documents and check a request against a payor’s requirements before it goes out, and payors use it to screen the requests they receive.
Federal and state rules set a limit on the payor side. A plan may use AI to assist, but a licensed clinician decides whether care is medically necessary, so a denial still comes from a person.
At a glance
- More than half of health plans and a quarter of providers use AI, per the 2025 CAQH Index.
- Medicare Advantage plans must base each medical necessity decision on the individual patient.
- California SB 1120 and Texas SB 815 bar AI from making medical necessity denials on its own.
- UnitedHealthcare’s post-acute denial rate rose from 8.7% in 2019 to 22.7% in 2022.
- Health plans committed to answer 80% of complete electronic approvals in real time in 2027.
“An extremely innovative and valuable service provided by true professionals!”Dexter M., RCM Lead, Chicago Speech TherapyLeave the document checks to Silna’s AI.Get started
Where AI works in a prior authorization
The 2025 CAQH Index found that more than half of health plans and a quarter of providers now use AI. Providers mostly use it for eligibility checks, note-taking and patient communications.
| Step | What AI can do for the practice | What AI can do for the payor |
|---|---|---|
| Checking whether approval is needed | Flags services on the payor’s approval list for the patient’s plan | No payor step yet |
| Gathering documents | Pulls notes and records from the EHR | No payor step yet |
| Building the request | Checks each document against the payor’s requirements | No payor step yet |
| Submitting | Sends the request and tracks its status | Receives and logs the request |
| Review | Tracks follow-up deadlines | Screens the request against coverage criteria |
| Decision | Flags a denial for staff to review and appeal | Assists, while a licensed clinician decides any denial |
On the payor side, a licensed clinician makes any denial. On the practice side, a person should check any request the software flags.
What a payor’s AI may decide on its own
On medical necessity, very little. The Medicare Advantage rules, two state laws and the AMA’s policy all keep the medical decision with a person.
| Rule | Who it covers | What it says about AI |
|---|---|---|
| CMS rule, 42 CFR 422.101(c) | Medicare Advantage plans | AI may assist, but each decision rests on the patient’s history, physician recommendations and clinical notes |
| California SB 1120 | Health plans and insurers in California | AI may not deny, delay or modify care based on medical necessity. A licensed physician or health care professional decides. |
| Texas SB 815 | Utilization review agents in Texas, from January 1, 2026 | An automated decision system may not make an adverse determination, wholly or partly |
| AMA policy, June 2023 | Insurers that use AI (a recommendation, not a law) | Physicians with relevant expertise review the decision, and someone examines the patient’s records before a denial |
CMS’s February 2024 guidance adds that a tool which decides coverage from a larger data set does not comply with the Medicare Advantage rule.
Original Medicare now has a pilot of its own. CMS’s WISeR Model adds prior authorization for 13 services in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. It uses AI and machine learning to review requests, and a licensed clinician makes every non-payment recommendation. Silna’s guide to the Medicare pilot program covers what it means for your requests.
State laws on AI in utilization review are still changing, so check your state’s current rule.
What denial data shows about payor automation
A US Senate Permanent Subcommittee on Investigations staff report (October 2024) found that UnitedHealthcare’s Medicare Advantage denial rate for post-acute care rose from 8.7% in 2019 to 22.7% in 2022. UnitedHealthcare had approved a “Machine Assisted Prior Authorization” program in April 2021. The report says human reviewers issued the final denials.
Physicians share the concern. In the AMA’s 2025 survey of 1,000 physicians, 60% said they are concerned that payors’ use of AI increases or will increase denial rates. The AMA uses the term augmented intelligence.
A denied request can still be appealed, and the appeal goes to a reviewer at the payor. Silna’s guide covers how to appeal a prior authorization denial.
What payors have promised for 2027
On June 23, 2025, health plans announced commitments with HHS to simplify prior authorization. Two of them bear on AI. In 2027, at least 80% of electronic prior authorization approvals that arrive with all needed clinical documentation are to be answered in real time, and medical professionals are to review every clinical denial.
The real-time target covers complete electronic requests only, so it doesn’t cover a request missing a document. Physicians are also doubtful. Only 33% in the AMA’s 2025 survey said the commitments will make a meaningful difference.
What to check before an AI tool touches your requests
Ask four things of any AI tool before it handles your requests:
- Does a person step in when the AI isn’t confident, and check a sample of what it submits?
- Does it apply each payor’s and specialty’s own documentation rules?
- Does it keep a record of what was sent, and when?
- Does it work inside the EHR or intake system your staff already use?
Silna’s comparison of the best prior authorization software reviews the main tools.
How Silna uses AI on prior authorizations
Silna’s Predictive Document Intelligence checks each document against payor- and specialty-specific requirements before a request is submitted, so a missing signature, credential or date is caught before the payor sees it.
Silna’s browser agents then submit requests on payor portals, which is how most payors take them. When an agent isn’t confident, it hands the request to one of Silna’s authorization specialists, who also review a sample of completed submissions.
Its Interoperability Agent works inside your EHR, intake system or CRM the way a trained employee would, pulling data in and pushing results out.
Silna’s AI-powered platform is backed by operators with years of healthcare experience. G2 reviewers rate Silna 4.8 out of 5 across 66 verified reviews.
Free your staff from re-checking every request by hand
Key terms
- Utilization review
- The process a payor uses to decide whether a requested service is medically necessary and covered.
- Medical necessity
- The standard a payor applies to decide whether a service is appropriate for the patient’s documented condition.
- Adverse determination
- A payor’s decision that a requested service does not meet its coverage or medical necessity criteria.
- Automated decision system
- Software that makes, or helps make, a decision such as approving or denying a request.
- WISeR Model
- A CMS pilot that adds prior authorization for selected Original Medicare services in six states, running from 2026 to 2031.
Frequently Asked Questions
Can a payor’s AI deny a prior authorization?
Not on its own under the rules covering most of these decisions. Medicare Advantage plans must base each medical necessity decision on the patient’s own records, California and Texas bar AI from making those denials, and WISeR leaves non-payment to licensed clinicians.
Will AI make prior authorization faster?
For complete electronic requests, health plans committed to answer at least 80% of approvals in real time in 2027. A request missing documentation falls outside that target.
Is WISeR an AI prior authorization program?
Partly. CMS says WISeR uses AI and machine learning to review 13 Original Medicare services in six states, and licensed clinicians make every non-payment recommendation.
How many providers use AI today?
A quarter, according to the 2025 CAQH Index, which found providers mostly use it for eligibility checks, note-taking and patient communications.
This article is general educational information and does not give medical, legal or insurance advice. Federal and state rules on AI in prior authorization change, so check the current rule before you rely on it.
About the author
Jeffrey Morelli
Jeffrey Morelli is the Co-Founder and CEO of Silna Health, the first Care Readiness Platform built to remove the administrative barriers that delay care. Silna automates benefit checks, eligibility, and prior authorizations across 1,000+ payors, and is backed by $27M from Accel and Bain Capital Ventures. Before Silna, Jeff spent a decade in San Francisco building and scaling products for highly regulated industries, including leading go-to-market at Truework (Series C, acquired by Checkr).
Last updated: November 2, 2026.