AI tools that help hospitals assign medical billing codes may be inflating what U.S. health insurers pay — and the early numbers are large.

The Blue Cross Blue Shield Association (BCBSA), which represents insurers covering roughly one in three Americans, has published an analysis finding that more intensive coding added an estimated $942 million to inpatient costs over two years. According to the association, the increase reflects more severe diagnoses being recorded without a corresponding increase in the treatments actually delivered.

The mechanism BCBSA describes is a variant of "upcoding." As hospitals adopt AI-assisted coding software, the tools scan patient records and suggest additional or more specific diagnoses — a legal and often clinically justified process that nonetheless drives up the severity level at which each case is billed. The association's comparison suggests hospitals using such tools billed more severely than facilities treating comparable patients without them.

Coverage of the report by The New York Times and others added further figures: costs for Blue Cross and Blue Shield plans rose by an estimated $653 million between 2024 and 2025, and hospitals received nearly $12,000 more per case on average.

Insurance-industry findings on this topic warrant a careful read. Insurers have a direct financial interest in arguing that hospitals overcharge them, and the BCBSA analysis is not a peer-reviewed study. Hospitals, for their part, argue that better documentation reflects genuinely better capture of conditions that were previously missed — information that can matter for patient care and for risk adjustment.

Still, the dispute points at a structural feature of the U.S. system that AI is now accelerating on both sides: payers and providers are locked in a coding arms race. Providers use algorithms to document and bill more completely; insurers use algorithms to detect and deny claims they consider unsupported. Each side describes the other's automation as the source of spiralling costs, and patients sit at the point where the two systems meet.

The BCBSA report is likely to become ammunition in that fight, and a signal that AI's effect on health spending is arriving faster than regulators and actuaries can measure it.

Sources