AI tools are driving up health insurance costs by nearly $1 billion, Blue Cross study finds
Artificial intelligence is helping doctors find more information in patient records, automate notes, and document care faster. But the benefits are increasingly coming with costs that few saw coming.
The AI boom has already been blamed for pushing up electricity demand and consuming vast amounts of water at data centers. Now another unexpected cost is emerging, this time inside the healthcare system. A new Blue Cross Blue Shield Association study points to another cost: higher medical spending.
AI-assisted documentation and coding contributed to nearly $1 billion in added costs for Blue Cross insurers over two years, according to the study released Thursday. The analysis found that providers increasingly documented secondary conditions that made patient cases appear more complex and led to higher reimbursement.
Between 2024 and 2025, those secondary diagnoses added about $653 million in costs for Blue Cross companies. The study found that more intensive coding and care classifications contributed $942 million in total added costs compared with 2023.
“AI tools are raising health spending for insurers by nearly $1 billion over two years, as providers bill for more severe patient care,” Reuters reported, citing a Blue Cross Blue Shield Association study released Thursday.
The technology behind that shift is already common across healthcare systems. Providers are using AI to scan medical records for missed conditions and deploying ambient scribes that listen to doctor-patient conversations and automatically draft clinical notes.
Those AI tools can save physicians time and surface details buried inside years of medical history. They can also change how a patient visit is coded.
A visit documented as medically more complex can qualify for a higher insurance payment. In hospitals, secondary and coexisting conditions can raise the reimbursement attached to a case.
That is where insurers are raising questions.
“If patients are truly sicker, we’d expect to see more treatment,” said Luke Chalker, senior vice president of product and data science at BCBSA.
More diagnoses, but not necessarily more treatment
BCBSA represents 31 independent Blue Cross Blue Shield companies covering more than 100 million people. Its study examined inpatient billing, focusing on cases where patients were admitted to hospitals or other medical facilities.
Among people undergoing major bowel surgery, diagnoses for certain secondary conditions rose sharply between the first quarter of 2023 and the fourth quarter of 2025. Partial intestinal blockages increased 55%, and diagnoses involving excess acid in the body rose 33%, according to the study.
The increase in diagnoses did not come with a comparable rise in treatment.
Dr. Razia Hashmi, vice president of clinical affairs at BCBSA, said more complex bowel surgery cases were not matched by higher treatment rates. Patients diagnosed with anemia, for example, did not receive more blood transfusions, a common treatment for serious red blood cell deficiency.
“The disconnect between diagnoses and treatment suggests that AI is identifying more billable conditions, not sicker patients,” said Chalker.
That distinction matters.
AI may be uncovering real conditions that doctors previously failed to document, making billing more accurate. Insurers see a different risk: software may be making it easier to capture every possible diagnosis that qualifies for a higher payment, without a matching change in the care patients receive.
Centene and other health insurers have already raised concerns that AI-assisted coding tools are contributing to aggressive reimbursement claims from healthcare providers.
The findings add another layer to the economics of AI adoption. Much of the conversation around AI in healthcare has focused on faster documentation, lower administrative workloads, and improved detection. The Blue Cross data points to a less obvious effect.
AI can make healthcare records more complete. It can also make them more expensive.
