Blue Cross Blue Shield Association, the trade group for Blue Cross and Blue Shield insurers across the country, just published an analysis with a specific number attached to a familiar complaint. Hospitals using AI tools to help write up patient charts, it says, added 942 million dollars in extra healthcare spending between 2023 and 2025. The group's language is blunt: patients are suddenly being documented as having far more complex conditions, but there is no sign the actual care delivered changed at all.
The mechanism is not mysterious. A new wave of AI note-taking tools, often called ambient scribes, listen to a doctor's conversation with a patient and generate a clinical note automatically. These notes tend to be longer and more detailed than what a busy doctor would type by hand. In hospital billing, more detail in a chart can justify a higher-paying billing code, even when the patient walked in with the same problem and left with the same treatment as before. Insurers call this upcoding. BCBSA estimates that roughly a fifth of a recent 9 percent rise in per-patient hospital costs traces back to this coding effect rather than to more or better care.
Hospitals push back hard on that story. Their trade group, the American Hospital Association, argues the numbers reflect a patient population that is genuinely older and sicker, and says AI is simply helping doctors document conditions that were always present but previously went unrecorded. Both explanations can be partly true at once, which is exactly why this fight is so hard to settle with data alone.
This is not a new argument, just a faster version of an old one. Medicare regulators have spent years tracking a similar pattern in Medicare Advantage, the privately run version of Medicare, where one government estimate pinned 40 billion dollars of a projected 84 billion dollars in 2025 overpayments on coding intensity rather than genuine illness. What is new is that AI is now pushing this same dynamic into everyday commercial insurance, and doing it faster than regulators or auditors can keep up.
Insurers are not innocent bystanders here. Reporting from health industry outlets shows large insurers are already deploying their own AI to catch and reverse these higher billing codes, and separately, insurers have faced lawsuits over AI systems used to deny claims outright, including a widely covered case against UnitedHealth's nH Predict algorithm. So the real picture is two industries pointing AI at each other, with hospitals trying to document their way to higher payments and insurers trying to compute their way out of paying them.
None of this happens in a vacuum for the people paying premiums. Milliman, one of the actuarial firms insurers rely on to set next year's prices, named AI as a cost driver in its most recent outlook for the first time. That means this fight is already baked into the premium increases employers and individuals will see.
The lesson here reaches past healthcare. Wherever one side profits from generating more detailed paperwork and the other side profits from catching it, count on both sides eventually deploying AI against each other. Insurance claims, warranty disputes, procurement audits, and expense reporting all share this structure. The businesses that come out ahead will be the ones using AI to be more accurate, not the ones using it to pad the numbers, because the other side is building a detector for exactly that.