Health insurers increasingly run claims and prior-authorization requests through algorithms that can approve, flag, or deny a request in seconds — work that used to take a human reviewer days. The technology speeds up routine approvals, but when the same software also drives denials, it can reject legitimate care faster than a patient or doctor can catch the mistake. That tension — speed versus accuracy, cost savings versus patient protection — is why AI claims review has become one of the most contested uses of AI in any industry.
How AI claims review actually works
A modern insurance claim or prior authorization request rarely reaches a human first. Optical character recognition and natural-language tools read the submitted forms and clinical notes, then a rules engine checks the diagnosis and procedure codes against the insurer’s coverage criteria. Predictive models compare the request to patterns in past claims — flagging ones that look like fraud, that don’t match a standard treatment pathway, or that fall outside an algorithmically-derived length-of-stay estimate. According to a 2023 survey by the National Association of Insurance Commissioners (NAIC), 84% of responding health insurers already use AI or predictive models somewhere in prior authorization or fraud detection. For routine, clearly-covered requests, this pipeline can issue an approval almost instantly. For anything ambiguous, the same pipeline can generate a denial letter — sometimes without a named clinician ever having reviewed the individual case.
Why the systems are under fire
The clearest example is a 2023 class-action lawsuit against UnitedHealth Group over nH Predict, a tool built by its subsidiary naviHealth to estimate how much post-acute care (nursing homes, rehab) an elderly patient should need. The plaintiffs allege the tool overrode treating physicians’ recommendations, and that patients who appealed a denial won reversal roughly nine times out of ten — while only a tiny fraction of denied patients ever filed an appeal. UnitedHealth disputes that the tool makes coverage decisions at all, saying it is a planning guide and that human medical directors make the final call. A federal judge has since ordered the company to turn over internal records on how the algorithm was used, and the case is proceeding through discovery. Similar lawsuits have targeted other large insurers. The underlying complaint is less about AI being wrong occasionally — every review process makes mistakes — and more about scale and opacity: a model that can issue thousands of denials a day, using boilerplate language keyed to diagnosis codes rather than an individual’s actual medical record.
What rights patients have now
Regulators have started to respond. The NAIC’s Model Bulletin on the Use of Artificial Intelligence Systems, adopted in December 2023 and since taken up by more than half of US states, requires insurers to maintain a written governance program for any AI system that affects coverage decisions and to let state regulators inspect how it was built, tested, and audited. California went further with the Physicians Make Decisions Act (SB 1120), effective January 1, 2025: any denial, delay, or reduction of care based on medical necessity must be reviewed by a licensed physician or other qualified clinician with relevant expertise, AI cannot be the sole basis for the decision, and the law sets hard deadlines — five business days for standard requests, 72 hours for urgent ones. Other states have introduced comparable bills. None of this bans AI from claims review; it requires that a human remain accountable for the final call, and it gives regulators a paper trail to check.
If a claim is denied, patients can generally ask the insurer whether AI was involved and request the name and specialty of the clinician who reviewed the case — insurers in states with these laws are required to be able to answer. Given how often algorithmic denials get reversed on appeal, filing that appeal is often worth the effort.
In the news
The tension between efficiency and oversight showed up directly this week, when Montefiore Medical Center laid off 12 nurses and shifted their insurance-review work to AI systems — a concrete example of hospitals and payers alike leaning on automation for exactly the kind of review process now under regulatory scrutiny.
FAQ
Can an insurer legally deny my claim using only AI, with no human review?
Increasingly, no — in states that have adopted laws like California’s Physicians Make Decisions Act, a licensed physician must review any denial based on medical necessity. Rules vary by state, so check your state’s insurance department for specifics.
How do I find out if AI was involved in my denial?
Ask your insurer directly, and check the denial letter for a named reviewing clinician. Some states now require insurers to disclose AI use on request.
Does AI review make insurance worse for patients?
Not uniformly — the same automation approves routine, clearly-covered claims almost instantly. The concern is concentrated in ambiguous or high-cost cases, where speed can come at the expense of an individualized review.
Sources: NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers (Dec. 2023); California SB 1120, the Physicians Make Decisions Act; KFF, “Regulation of AI in Prior Authorization and Claims Review” (2026); CBS News and Healthcare Finance News reporting on the UnitedHealth nH Predict litigation.