Medical AI drives nearly $1 billion in extra costs for Blue Cross Blue Shield insurers

A study by the Blue Cross Blue Shield Association (BCBSA), released Thursday, estimates that the use of artificial intelligence tools in clinical documentation has driven up health spending for its insurers by nearly $1 billion over two years. According to the organization, health providers are increasingly recording secondary conditions — ailments stemming from an illness other than the one that prompted the visit — thanks to AI tools that scan patient records or use 'ambient scribes', systems that passively listen in on conversations with patients and automatically draft the medical notes.

The study, which analyzed hospital billing between 2023 and 2025, attributes $653 million of the cost increase directly to the rise in billing for secondary conditions during the 2024-2025 period. All told, care documented as more complex or intensive added $942 million in extra costs for BCBS companies over those two years, compared with the 2023 baseline.

One specific example cited in the report: among patients undergoing major bowel surgery, the recording of secondary conditions such as partial intestinal blockages and acid overloads in the body rose by 55% and 33%, respectively, between the first quarter of 2023 and the fourth quarter of 2025.

What raises BCBSA's suspicions is not just the increase in diagnoses, but that they are not accompanied by more treatment. Luke Chalker, the association's senior vice president of product and data science, sums it up this way: if patients were genuinely sicker, you would expect to see more associated treatment. Yet Razia Hashmi, BCBSA's vice president of clinical affairs, notes that in bowel disease surgery cases the greater diagnosed complexity did not come with higher treatment rates. One cited instance: when anemia was diagnosed, blood transfusions — the standard treatment for that condition — did not increase. For Chalker, that disconnect between diagnoses and treatments suggests AI is identifying more billable conditions, not sicker patients.

Medical visits documented as involving more complex care allow providers to charge insurers higher rates, and in hospital settings secondary or coexisting conditions are typically classified as higher complexity. The report adds to concerns already voiced by other insurers: Centene, for example, has previously noted that health systems' use of AI tools has led to aggressive or inappropriate reimbursement payments.

BCBSA brings together a network of 31 independent health insurance companies that together cover more than 100 million people in the United States, giving this analysis of hospital billing considerable weight within the sector. The case illustrates a tension likely to recur in other areas of AI-driven healthcare: the very tools that promise to ease doctors' administrative burden and improve clinical documentation can also, without any explicit malicious intent, systematically inflate the recorded complexity of each case and, with it, the cost ultimately borne by the insurance system and, in the end, by patients.