You’re holding a letter that says your claim was denied. Maybe it says “not medically necessary” and nothing else. You call to ask who reviewed it, and nobody on the phone can give you a name. If it feels like no doctor ever actually looked at your specific situation, you might be right — and if you’re asking whether an algorithm denied your health insurance claim instead of a person, this guide is built to help you check.
If your denial letter uses vague language, names no reviewing doctor, and leans entirely on a billing code, there’s a real chance an algorithm — not a physician — made the call. And depending on your plan, you may have a specific legal right to make the insurer prove otherwise.
And here’s the tool almost nobody uses: in many cases, you can legally demand that the insurer show its work.
Decode Your Denial Letter
Answer four quick questions about your denial letter. This won’t tell you whether an algorithm denied your claim — no online tool honestly can. It will tell you how many red flags are present and which specific legal lever applies to your plan.
This tool never concludes that your claim was denied by AI. It scores checkable red flags and routes you to the legal right that matches your plan type. Nothing you enter is saved, stored, or sent anywhere — it’s calculated in your browser only.
First, What Kind of Plan Do You Have?
This is the single most important question in this whole guide, and it’s the one almost every other article on this topic skips. The legal tools below only work if you point them at the right target. Your rights change completely depending on how you get your coverage.
| Plan type | Governing framework | Key right | How to invoke it |
|---|---|---|---|
| Employer group plan (private-sector job) | ERISA — 29 CFR §2560.503-1 | Free copies of the full claim file, including anything relied on or generated to make the decision | Written request citing §2560.503-1(h)(2)(iii) and (m)(8); see the ERISA section below |
| Medicare Advantage | CMS-0057-F, plus standard Medicare Advantage appeal rights | Specific clinical criteria in the denial notice; 7-day standard / 72-hour expedited decisions; automatic referral to an Independent Review Entity if upheld | Cite CMS-0057-F in your appeal; use the expedited process if care is ongoing |
| Medicaid managed care | 42 CFR Part 438, plus CMS-0057-F for the managed-care organization’s prior-authorization process | Internal MCO appeal, plus a state Medicaid fair hearing | Contact your state Medicaid agency’s fair hearing office; ask your MCO for the CMS-0057-F-required denial reason |
| ACA marketplace / individual plan (QHP) | ACA Section 2719 external review, plus CMS-0057-F for QHPs on federally facilitated exchanges | Internal appeal, then independent external review by a neutral reviewer | Request external review through HealthCare.gov or your state-based exchange after internal appeal |
| Government or church employer plan | Usually exempt from ERISA; governed by state insurance law (if fully insured) and the plan document | Varies — some states extend similar AI-review or disclosure requirements to state-regulated plans | Contact your plan administrator and your state insurance department to find out what applies |
A wrinkle worth knowing: if your employer “self-funds” its health plan — common at large employers — your state’s AI-review law (like California’s SB 1120) may not apply to you at all, because federal ERISA law generally preempts state insurance regulation of self-funded plans. Your ERISA claim-file right still applies regardless. Ask your HR or benefits department whether your plan is self-funded or fully insured; it’s usually printed in your Summary Plan Description.
Not sure which category you’re in? Ask your HR department (for job-based coverage), check your insurance card for “Medicare Advantage” or your plan’s Medicaid/marketplace name, or look at your Summary Plan Description. Everything below is organized so you can jump to the section that matches your plan.
Red Flags: Decode Your Denial Letter
This isn’t proof — it’s a reason to ask. None of the signs below can tell you, by themselves, that an algorithm denied your claim without a real physician looking at your case. They’re checkable signs your insurance denial was automated, not a diagnosis of your specific letter. Insurers use software throughout their process; the question is how much of the actual decision a human being made.
| What it says | What it might mean | Your next step |
|---|---|---|
| “Does not meet medical necessity” or “does not meet clinical guidelines,” with no clinical criteria cited | No specific standard was applied, or a generic template reason was used | Ask for the exact clinical criteria and guideline applied to your case |
| The denial references only a billing or diagnosis code | Possibly a code-matching automated screen rather than individualized review of your records | Ask whether an automated or algorithmic tool was used, and request its output |
| No named reviewing physician, or only a generic “medical director” title — an insurance denial letter with no doctor’s signature | The reviewer’s identity may not be individualized, or review may have leaned heavily on automated output | Request the reviewer’s name, license, and specialty in writing |
| The denial arrived unusually fast for a complex claim | Possible batch or automated processing — speed alone isn’t proof, but it’s worth combining with other flags | Note the turnaround time in your appeal; combine with other red flags below |
| The reviewer’s specialty doesn’t match your condition | The reviewer may lack the relevant expertise to evaluate your case | Request a peer-to-peer review with someone in the same or similar specialty |
| An internal policy number is cited with no plain-language explanation | Boilerplate policy citation without case-specific reasoning | Request a copy of the specific policy and how it was applied to your case |
Score your letter: three tiers
A named clinician with credentials, specific clinical criteria tied to your records, and reasoning that engages with your actual case.
Generic boilerplate language, no named reviewer, and reasoning that cites only a code or an internal policy number.
Key information is simply missing. Your job here is to force clarity using the state and federal tools in this guide.
The Lawsuits Behind This
These are allegations in active, unresolved litigation — not proven facts, and not a ruling that any insurer broke the law. Case status moves quickly; verify the current docket before relying on specifics. As of September 2026:
UnitedHealth: Estate of Lokken v. UnitedHealth Group
Filed in the District of Minnesota in November 2023, this class action centers on nH Predict, a predictive AI tool from UnitedHealth’s Optum subsidiary naviHealth. The plaintiffs allege UnitedHealth used the tool to override physicians’ clinical judgment and cut off post-acute care for Medicare Advantage members prematurely, and cite a high reversal rate on internal appeal. In March 2026, a federal magistrate judge ordered UnitedHealth to turn over an expansive set of internal documents on its AI-driven claims process, some reaching back to 2017. UnitedHealth’s Optum unit disputes the characterization, maintaining nH Predict is a care-support tool rather than a claims-adjudication tool, and that physicians make the actual medical-necessity calls. No court has ruled on the merits of these allegations.
Humana: Barrows v. Humana
A related class action filed in the Western District of Kentucky in December 2023 makes similar allegations against Humana over its use of the same nH Predict tool for Medicare Advantage post-acute care denials. A federal judge allowed the case to proceed past a jurisdictional challenge in 2025, and the docket remains active into 2026. Humana has not commented on the specific allegations, citing the pending litigation.
Cigna: Kisting-Leung v. Cigna Corp.
Filed in the Eastern District of California in July 2023, this case centers on Cigna’s “PxDx” claims-review software. The complaint cites a ProPublica investigation reporting that Cigna doctors could reject a batch of claims without individually reading patient files, and that reviews took an average of roughly 1.2 seconds per claim. A March 2025 court ruling narrowed some of the plaintiffs’ claims but allowed a fiduciary-duty claim and a California unfair-competition claim to proceed; the case remains active. Cigna disputes that PxDx uses artificial intelligence, and says the review happens after treatment is provided, so it does not deny care outright — only reimbursement in some instances.
The throughline: none of these cases has produced a final ruling that any insurer’s algorithm illegally denied care. What they’ve produced, so far, is enough evidence for federal judges to let the claims proceed — which is itself a meaningful signal that the underlying practices deserve scrutiny.
Your Legal Shield: State and Federal Rules
Your right to clinical criteria in an insurance appeal comes from a mix of federal rules and state laws, and which one protects you depends on your plan type from the router above. Here’s the current landscape.
Medicare Advantage: CMS-0057-F
Effective January 1, 2026, the CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F) requires Medicare Advantage plans, Medicaid and CHIP managed care plans, and ACA marketplace plan issuers on the federally facilitated exchanges to decide standard prior-authorization requests within 7 calendar days, expedited requests within 72 hours, and to give a specific reason for every denial — a code alone no longer satisfies the requirement. It does not cover traditional Medicare, standalone Part D, or most employer-sponsored plans.
State laws requiring human clinical review
State legislatures have moved unusually fast on this issue. Several states enacted new requirements within months of each other in 2025 and 2026:
| State | Law | What it requires | Effective date |
|---|---|---|---|
| California | SB 1120, “Physicians Make Decisions Act” | A licensed physician or qualified clinician, not AI alone, must decide medical-necessity denials | January 1, 2025 |
| Illinois | HB 2472 | If an algorithmic process is used in utilization review, only a “clinical peer” may issue the adverse determination | January 1, 2025 |
| Maryland | HB 820 | Flat prohibition on AI as the basis for a denial (not just a “sole basis” standard) | October 1, 2025 |
| Texas | SB 815 | Prohibits AI as the sole basis for a medical-necessity denial; gives the Texas Department of Insurance audit authority | Signed June 2025; human-review mandate applies to plan years beginning January 1, 2026 |
| Indiana | HB 1271 | Bars insurers from downcoding a claim based on AI output alone, without human review of the medical record | July 1, 2026 |
| Washington | SB 5395 | Prohibits sole reliance on AI to deny, delay, or limit care; requires carriers to report AI-assisted denial counts | June 11, 2026 |
| Alabama | SB 63 | Restricts AI use in utilization review determinations | October 1, 2026 |
This list is not exhaustive and it is growing. Separately, a majority of states have adopted the National Association of Insurance Commissioners’ AI Model Bulletin — a supervisory framework that pushes insurers toward documented AI governance. It’s a useful backdrop, but it’s a regulatory tool for state examiners, not a specific right you can personally invoke in an appeal. Check your own state insurance department’s website for the current list.
The Peer-to-Peer Review Sham
A peer-to-peer review is supposed to be a conversation between your treating doctor and a reviewer with comparable training, about your specific case. In practice, physicians and researchers have documented a pattern where that’s often not what happens.
An American Medical Association survey reported that only 16% of physicians said the insurer’s “peer” reviewer often or always had the right qualifications for the case, and physicians commonly describe talking to reviewers from an entirely different specialty than the one at issue. A ProPublica investigation documented a specific, verifiable case: a Cigna member’s claim for a vitamin D test was denied in a letter signed by a named Cigna medical director, and the denial was later reversed after an independent, outside reviewer examined the same records and found the test was justified. That’s one documented example, not proof of a universal practice — but it’s the kind of pattern this guide asks you to check for in your own case.
What a legitimate review looks like
- The reviewer discloses their name, license, and specialty
- The reviewer’s specialty matches or is closely related to your condition
- The reviewer engages with your specific records and your doctor’s rationale
- You get real advance notice and an actual two-way conversation
Red flags of a rubber stamp
- The reviewer’s specialty has nothing to do with your condition
- The call reads from a script and won’t engage with your specific case
- The call is over in minutes, with no discussion of your doctor’s documentation
- The insurer won’t disclose the reviewer’s name or credentials up front
The ERISA Weapon: Request the Full File
You can legally demand the file. If your coverage comes through a private-sector employer, federal law gives you the right to request — free of charge — everything the insurer relied on, submitted, considered, or generated in deciding your claim.
Scope, stated plainly: this right comes from ERISA and applies only to private-sector, employer-sponsored group health plans. It does not apply to Medicare, Medicaid, ACA marketplace plans, or most government and church employer plans — those have the separate rights covered in the section above. If you’re not sure which category you’re in, go back to the plan-type router before using this section.
The right itself comes from 29 CFR §2560.503-1(h)(2)(iii) and (m)(8). A document is “relevant” to your claim — and must be disclosed on request — if it was relied on in the decision, or simply submitted, considered, or generated while the decision was being made, whether or not the plan ultimately relied on it. In plain terms: if an algorithmic or predictive tool produced any output as part of your case, that output is very likely covered.
How to Request Your Insurance Claim File Under ERISA
Put your request in writing — email, the insurer’s appeal portal, or certified mail — and keep a copy. A workable version:
“I am requesting, under 29 CFR §2560.503-1(h)(2)(iii) and (m)(8), copies of all documents, records, and other information relevant to my claim. This includes any algorithmic, predictive, or automated tool output that was used, considered, or generated in connection with this determination, and the name, license, and specialty of anyone who reviewed my claim.”
There’s no single fixed deadline written into the regulation for this specific document request, separate from the deadlines that apply to deciding your claim itself. Make the request as early as possible — ideally alongside your written appeal — and treat a plan’s silence or delay as something to escalate, either by following up in writing or by asking a benefits attorney or your state insurance department how to proceed.
Fighting Fire With Fire: Using AI to Write Your Appeal
Read this before you draft anything: an AI tool can help you organize and phrase an appeal letter. It does not know your specific plan documents, your state’s current law, or your medical facts. Every citation, every fact, and every clinical detail in a drafted appeal needs to be checked by you — or your doctor, or a lawyer — before you send it. This is not a substitute for legal or medical advice.
Using an AI tool to help write an appeal is a real, growing practice, and there’s nothing wrong with it as a starting point. Here’s a realistic breakdown of what it can and can’t do.
What it can help with
- Organizing your denial letter, plan documents, and medical records into a clear timeline
- Drafting a structured letter once you tell it which legal framework applies to your plan (from the section above)
- Suggesting phrasing that ties your specific medical records to the insurer’s stated reason for denial
- Drafting a peer-to-peer review request or a follow-up letter
What it can’t do
- Verify your specific plan’s terms, exclusions, or coverage rules
- Know your state’s current law with certainty — this is one of the fastest-moving areas of legislation in the country right now
- Confirm that any citation, case name, or regulation number it produces is accurate or current
- Make a medical judgment about your case, or replace a treating doctor’s letter of medical necessity
Practical order of operations: gather your denial letter, plan documents, and medical records first. Tell the tool which plan-type framework applies. Let it draft — not decide. Then verify every fact, citation, and clinical detail against your own documents, or with your doctor, before you send anything. A letter of medical necessity from your treating physician should still be the core of your appeal, not the AI-drafted letter itself.
One more time, because it matters: check every fact before it goes out the door. An appeal built on a wrong citation can weaken your case instead of strengthening it.
What This Isn’t
This guide is about one specific, modern cause of denials — algorithmic and AI-assisted review — and the counter-tactics it calls for. For the general appeals process, deadlines, and paperwork that apply to any denied claim, see How to File a Health Insurance Appeal for a Denied Claim.
FAQ
- Can I ask my insurer if AI reviewed my claim?
- Yes. You can ask directly, and depending on your plan type, CMS-0057-F or a state law may require the insurer to disclose it.
- Does ERISA apply to my plan?
- Only if you have private-sector, employer-sponsored group coverage. Ask HR whether your plan is ERISA-covered; government and church employer plans usually are not.
- What is CMS-0057-F?
- A federal rule requiring Medicare Advantage, Medicaid and CHIP managed care, and ACA marketplace plan issuers on the federal exchange to give specific denial reasons and meet faster decision timelines, effective January 1, 2026.
- How do I request my insurance claim file?
- If you have an ERISA-covered plan, put your request in writing citing 29 CFR §2560.503-1(h)(2)(iii) and (m)(8). See the ERISA section above for wording.
- Is it illegal for an insurer to use AI to deny my claim?
- Not automatically. A growing number of states require human review of any AI-assisted denial, but the rules vary by state and plan type — check the legal shield section above.
- What if my denial letter doesn’t name a doctor?
- That’s a red flag, not proof of anything on its own. Request the reviewer’s name, license, and specialty in writing.
- How long do I have to appeal?
- For ERISA group health plans, generally at least 180 days from the denial. Medicare, Medicaid, and marketplace timelines differ — check your specific denial letter.
- Can an algorithm legally deny healthcare coverage?
- Not entirely on its own, in a growing number of states and plan types. A licensed clinician generally has to make or review the final medical-necessity decision — see the plan-type router and legal shield sections above for exactly which rule applies to you.
- Can a non-specialist doctor deny my claim?
- It happens, and it’s one of the documented complaints about peer-to-peer review qualifications — reviewers sometimes lack the same specialty as your treating doctor. It isn’t automatically illegal, but you can request a reviewer with matching expertise. See the peer-to-peer section above.
- What should I do if my health insurance claim is denied?
- Start with the closing checklist just below: identify your plan type, request your claim file if ERISA applies, ask for the reviewer’s name and credentials, and cite whichever state or federal rule fits your plan.
- Can I request a peer-to-peer review?
- In many cases, yes. Ask specifically for a reviewer in the same or a closely related specialty as your treating doctor.
- What if I have Medicare Advantage and my rehab or nursing care was cut short?
- Cite CMS-0057-F and ask for the specific clinical criteria used. Use the expedited appeal process if care is ongoing, and consider mentioning the pending nH Predict litigation if that tool was involved.
- What if my employer self-funds its health plan?
- It’s typically still ERISA-covered, so your claim-file right still applies — even though your state’s AI-specific insurance law might not, due to ERISA preemption.
- Should I use AI to write my appeal?
- It can help you organize and draft. Never send anything without verifying every fact and citation yourself, or with your doctor or a lawyer.
- Can a denial really get reversed?
- Yes. Reported reversal rates in the pending litigation described above, and outside independent-review data generally, suggest many initial denials don’t hold up under further review.
- What is “downcoding,” and does it relate to this?
- Downcoding is paying a claim at a lower level than what was billed. Some states, including Indiana, now separately regulate AI-driven downcoding alongside outright denials.
- Where can I find my state’s specific law?
- Check your state insurance department or commissioner’s office directly — new laws are being enacted quickly, and this guide’s state list won’t stay current forever.
Sources
- ArentFox Schiff, “Federal Court Orders Broad Discovery Against UHC in AI Coverage Denial Lawsuit” (2026) — afslaw.com
- ProPublica, “How Cigna Saves Millions by Having Its Doctors Reject Claims Without Reading Them” — propublica.org
- Georgetown Health Care Litigation Tracker, “Barrows et al. v. Humana Inc.” — litigationtracker.law.georgetown.edu
- 29 CFR §2560.503-1, Claims procedure — eCFR
- CMS, “CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F)” fact sheet — cms.gov
- California SB 1120 (2023–2024), official bill text — leginfo.legislature.ca.gov
- Texas SB 815 (89th Legislature), official bill analysis — capitol.texas.gov
This article is for general education only and is not legal or medical advice. The litigation described above involves unresolved allegations, not proven findings of fact; case status can change at any time. State and federal rules vary by plan type and change frequently — verify current requirements with your state insurance department or a licensed attorney before relying on them. If you’re dealing with a denial that affects your health, consider consulting a licensed attorney or a patient advocate about your specific situation.
Last updated:
Daniel Hayes is the founder and sole researcher at AdvoraHQ. He covers U.S. personal finance, insurance, and consumer law — working directly from IRS publications, federal and state statutes, court opinions, and SEC filings rather than secondary summaries. His focus is the gap between what readers think they know and what the source documents actually say. Daniel is not a licensed attorney, CPA, or financial advisor; his articles are educational and not personalized advice. Reach him at Daniel.Hayes@advorahq.com.
