Appealing AI-Involved Denials: Disclosure Rights and a Behavioral Health Playbook
How behavioral health teams should appeal denials where payer AI may have been involved: 2026 state disclosure rules, the decision-record request, evidence-first appeal structure, and regulator escalation.
On this page: Direct answer
Direct answer
AI insurance denial appeal: what operators need to know
How behavioral health teams should appeal denials where payer AI may have been involved: 2026 state disclosure rules, the decision-record request, evidence-first appeal structure, and regulator escalation. AI-assisted review is permitted in most markets; a growing set of states requires a human decision-maker and disclosure.
Payers increasingly use algorithms and AI inside utilization review — to triage requests, suggest determinations, and draft rationales. States have responded. California's SB 1120 prohibits an algorithm from being the sole basis of a medical-necessity denial and keeps the determination with a licensed professional. A 2025–2026 wave of state laws imposes similar human-decision and disclosure requirements, and Maryland now requires insurers to report adverse decisions quarterly, including whether AI or another software tool was used. In traditional Medicare, the CMS WISeR model is testing technology-assisted prior authorization for selected Part B services in six states from 2026 through 2031, with licensed clinicians reviewing non-affirmed requests.
For a behavioral health appeals team, the operational question is not whether AI touched the determination — you often cannot prove that from the notice. It is how to build appeals that succeed regardless, and how to use disclosure and human-review rights where they exist. Published analyses of Medicare Advantage data show that most appealed denials are overturned, which means the leverage is real. This guide is operational, not legal advice; confirm current law for each plan and state with qualified counsel.
Key takeaways
The short version
- AI-assisted review is permitted in most markets; a growing set of states requires a human decision-maker and disclosure.
- Ask in writing for the criteria used, the reviewer's identity and credentials, and whether an automated tool contributed.
- Win the appeal on criteria and evidence — the technology argument belongs in the decision-record request and the regulator complaint.
- Record determination speed, template language, and repeat rationales as case metadata, not accusations.
- Hold your own side to the same standard: AI may draft, but a named human approves everything you submit.
Take the template with you
Free to copy · no email required
Paste into a written reconsideration or records request to establish the criteria, the reviewer, and the role of any automated tool.
Regarding the adverse determination dated [DATE] for [MEMBER ID / CASE NUMBER], please identify in writing: (1) the specific clinical criteria or guideline relied upon, including the title, version, and effective date; (2) the name, licensure, and specialty of each individual who made or approved the determination; (3) whether any algorithm, artificial intelligence system, or automated decision-support tool was used at any stage of the review, and if so, its role in the determination; and (4) an itemized list of the clinical information reviewed. Please provide this information within the timeframe required by applicable law and plan documents. This request supplements, and does not replace, our pending appeal.
1. Know the 2026 rules of the game
One structural caveat governs everything above: state insurance laws generally regulate fully insured products. Self-funded employer plans under ERISA follow federal rules, so the same denial from the same administrator can carry different rights depending on plan funding. Capture funding type during benefits verification — it determines which arguments are available at appeal time.
| Layer | What it says | Where it applies |
|---|---|---|
| California SB 1120 | An algorithm cannot be the sole basis of a medical-necessity denial; a licensed professional makes the determination | California-regulated plans |
| 2025–2026 state wave | Multiple states prohibit AI-only adverse determinations or add disclosure and reporting duties | Varies by state; confirm the current statute |
| Maryland reporting | Insurers report adverse decisions quarterly, including whether AI or software tools were used, effective June 1, 2026 | Maryland-regulated insurers |
| CMS WISeR model | Technology-assisted prior authorization for selected Part B services, 2026–2031, with licensed-clinician review of non-affirmed requests | Traditional Medicare in AZ, NJ, OH, OK, TX, WA for listed services |
| CMS-0057-F | Impacted payers must provide specific denial reasons and meet decision timeframes | MA, Medicaid/CHIP managed care, and Exchange plans in scope |
2. Record the signals of automated review — without overclaiming
- Adverse determinations returned faster than a plausible clinical review of the submitted record
- Identical rationale language across different patients, diagnoses, or levels of care
- Denial reasons that mismatch the submitted documentation — citing an absent record that was included, or criteria for a different service
- Volume patterns: clusters of denials on one code or level of care that began on an identifiable date
- Reviewer fields that are blank, generic, or show credentials mismatched to behavioral health
3. Request the decision record
- 01
Name the criteria
Ask for the specific clinical criteria or guideline used, including the version and effective date. Payers applying behavioral health criteria must be able to identify them; the answer determines what your appeal argues against.
- 02
Identify the reviewer
Request the name, licensure, and specialty of each person who made or approved the adverse determination. Many states require that behavioral health determinations be made by a matched or qualified specialty reviewer.
- 03
Ask the AI question directly
Ask whether any algorithm, artificial intelligence, or automated decision-support tool was used at any stage of the review, and what role it played. Where state law restricts AI-only determinations or requires disclosure, cite it; where it does not, the documented non-answer is still useful.
- 04
Request the reviewed record
Ask what clinical information was actually considered. Mismatches between what you submitted and what was reviewed are among the strongest appeal facts available.
- 05
Send it in writing, on the clock
Route the request through the channel the notice specifies, within the appeal window, and log the response deadline as a case date.
4. Build the appeal on criteria, not on the technology
- 01
Map every denial reason to evidence
For each stated reason, attach the record excerpt that answers it — an exhibit per deficiency. An appeal organized reason-by-reason is easier for the next human reviewer to grant.
- 02
Write to the named criterion
Argue the payer's own criterion, element by element, in the clinical language of the record. Generic medical-necessity prose loses to criterion-mapped evidence.
- 03
Correct the reviewed record
Where the determination mischaracterized or missed submitted documentation, say so plainly, cite the original submission date, and resubmit the document as an exhibit.
- 04
Request a matched peer-to-peer
Ask for a peer-to-peer with a behavioral health clinician of the relevant specialty before the internal appeal concludes, and prepare it with the same criterion map.
- 05
Keep a named human on your side
AI can draft the appeal scaffold; a named clinician or appeals lead reviews and signs it. That is the same human-review standard you are asking the payer to meet.
5. Escalate patterns to regulators and contracts
- Exhaust internal appeal levels on the clock, then use external review rights — federal or state, depending on the plan
- File a state insurance-department complaint where a determination appears to violate an AI-review or reviewer-credential law, attaching the decision-record request and any non-response
- Aggregate the case metadata — speed, template rationales, overturn rates by payer and service — and bring it to joint operating committees and contract renewals
- In WISeR states, track CMS model materials and payer notices for the listed services so Medicare cases are routed with the model's timelines in mind
- Never let pattern arguments replace per-case deadlines: every case still needs its appeal filed on time
Common questions
Answers before you build.
How do I know if AI denied my claim?+
Usually you cannot tell from the notice alone. Disclosure and reporting laws are expanding — Maryland requires insurers to report whether AI was used in adverse decisions, and several states require human decision-makers — but in most markets the practical move is to ask in writing for the decision record and log the response.
Are AI-generated denials illegal?+
Not generally. Several states prohibit AI from being the sole basis of a medical-necessity denial and require a licensed professional to make the determination, and some add disclosure duties. AI-assisted review with a human decision-maker remains permitted in most markets. Confirm the current law for the specific plan and state.
Does Medicare use AI for prior authorization?+
Traditional Medicare historically used little prior authorization. The CMS WISeR model, running 2026 through 2031, tests technology-assisted review for selected Part B services in six states, with CMS stating that non-affirmed determinations are reviewed by licensed clinicians. Medicare Advantage plans run their own utilization management subject to CMS rules.
Should the appeal letter argue that AI made the decision?+
Lead with clinical evidence mapped to the payer's named criteria — that is what overturns denials. Raise decision-process concerns where you have a concrete hook: a state law on AI-only determinations, a reviewer-credential requirement, or a documented mismatch between the submitted and reviewed record. Broader pattern concerns belong in a regulator complaint.
Practical closeout
Use this operator checklist.
- AI-assisted review is permitted in most markets; a growing set of states requires a human decision-maker and disclosure.
- Ask in writing for the criteria used, the reviewer's identity and credentials, and whether an automated tool contributed.
- Win the appeal on criteria and evidence — the technology argument belongs in the decision-record request and the regulator complaint.
- Record determination speed, template language, and repeat rationales as case metadata, not accusations.
- Hold your own side to the same standard: AI may draft, but a named human approves everything you submit.
Continue through the cluster
Verified customer case studies are added only with customer permission and supporting evidence; none is implied by these operational examples.
Sources & methodology
Trace the operational claims.
Marsa Health Editorial reviewed the primary and research sources below on July 28, 2026. We translate them into workflow controls, distinguish proposals from final rules, and flag where plan, program, state, contract, or clinical requirements vary.
- 01SB 1120: Health care coverage: utilization review California Legislative InformationCalifornia statute limiting the role of algorithms and artificial intelligence in medical-necessity determinations and requiring licensed-professional review. Verify current text and applicability.Accessed or rechecked July 28, 2026
- 02States Continue Efforts to Regulate AI in Healthcare: A Review of Legislation Passed in 2026 Holland & KnightLaw-firm survey of 2025–2026 state legislation governing payer use of AI in utilization review and adverse determinations. Secondary source; confirm each state's current law with counsel.Accessed or rechecked July 28, 2026
- 03WISeR (Wasteful and Inappropriate Service Reduction) Model Centers for Medicare & Medicaid ServicesOfficial CMS Innovation Center page for the 2026–2031 WISeR model testing technology-assisted prior authorization for selected Part B services in six states, with licensed-clinician review of non-affirmed decisions.Accessed or rechecked July 28, 2026
- 04Medicare Advantage prior authorization determinations in 2024 KFFAnalysis of CMS data on Medicare Advantage prior authorization denials and appeals.Accessed or rechecked July 28, 2026
- 05How to appeal an insurance company decision Centers for Medicare & Medicaid ServicesFederal overview of internal appeals, external review, notices, and general appeal timing. Plan and state rules can differ.Accessed or rechecked July 28, 2026
- 06External Appeals Centers for Medicare & Medicaid ServicesFederal external-review process and consumer protections.Accessed or rechecked July 28, 2026
- 07CMS Interoperability and Prior Authorization Final Rule CMS-0057-F Centers for Medicare & Medicaid ServicesCurrent implementation dates, decision timeframes, denial-reason requirements, metrics, and API provisions for impacted payers.Accessed or rechecked July 28, 2026
Organizational author. Editorial review covers source accuracy, search intent, workflow boundaries, and human-oversight requirements. This material is educational and does not provide clinical, legal, coding, or coverage advice.
No named clinical or legal expert reviewer is attributed to this version. Marsa Health does not invent reviewer credentials.
Read our editorial methodRevision history
What changed and when
July 28, 2026
Initial publication, source review, and operational editing.