Key Takeaways
- The Blue Cross Blue Shield Association (BCBSA) attributes $942 million in added healthcare spending across 2023–2024 claims to hospitals’ use of AI-enabled medical coding tools, arguing higher payments were made for “similar care.”
- The American Hospital Association (AHA) formally rejects the finding, contending that patients are genuinely sicker and that improved coding reflects recovery from historical under-documentation rather than artificial inflation.
- The dispute centers on “coding intensity”, AI documentation tools that surface more diagnoses and higher-acuity billing levels, not on allegations of fraud.
- The claim lands amid a projected 9% medical cost trend for 2027, the steepest in 17 years, with AI-enabled coding named among the drivers alongside GLP-1 drug costs.
- Neither side has produced independently replicated causal evidence, leaving the $942 million figure a contested claim rather than a confirmed finding.
Insurer Group Points to AI Coding as Cost Driver
On September 24, 2026, the Blue Cross Blue Shield Association released an analysis attributing $942 million in added healthcare spending over two years to hospitals’ use of AI-enabled medical coding tools, a finding the American Hospital Association immediately and formally rejected.
The BCBSA analysis, covering 2023 and 2024 claims, concluded that health plans paid more for “similar care,” with the association pointing to AI documentation and coding software as the mechanism. The figure has been widely reported as “nearly $1 billion.” The AHA answered with a formal rebuttal on October 5, 2026, following an earlier fact sheet published August 26, 2026, framing the dispute as hospitals treating sicker patients versus insurers seeking to suppress legitimate payment.
The standoff has become the most visible flashpoint in a broader argument over who bears responsibility for rising medical costs as AI spreads through both clinical and administrative operations.
Key Facts
- BCBSA analysis: Covers 2023–2024 claims; alleges higher payments for “similar care.”
- AHA response: Rebuttal published October 5, 2026; fact sheet August 26, 2026.
- Vendor marketing: STAT News (August 20, 2025) reported that scribe vendors explicitly market revenue uplift, not merely time savings.
- Named health systems: Fierce Healthcare (February 17, 2026) reported that Rush, McLeod Health, and FMOL Health confirmed revenue gains from the Suki AI scribe.
- Mainstream coverage: STAT (April 8, 2026) summarized the impasse: “Everyone agrees AI scribes are increasing health care costs. No one agrees what to do about it.”
- Macro backdrop: PwC projects a 9% medical cost trend for 2027, the highest in 17 years (Fortune, June 12, 2026). ACA marketplace premiums are proposed to rise roughly 14% for 2027, while the Medicare Advantage 2027 rate increase was finalized at 2.48%.
- Scale check: National health expenditure runs roughly $5 trillion annually, against which $942 million is a directional signal, not an absolute-burden argument.
What “Coding Intensity” Actually Means
The technical heart of the dispute is coding intensity, the volume and acuity of diagnoses and billing codes attached to a patient encounter. AI documentation and coding tools, including ambient clinical scribes, computer-assisted coding, and clinical documentation improvement software, are designed to surface more diagnoses and support higher-acuity evaluation-and-management (E/M) levels.
BCBSA’s core argument is that AI has lowered the cost of documenting billable complexity, decoupling payment from actual clinical change. The association is not alleging fraud. It is making a volume-and-documentation argument: that the marginal cost of capturing a billable diagnosis has fallen sharply, and payment has followed.
Evidence that the mechanism is designed rather than accidental is substantial. STAT News reported in August 2025 that scribe vendors market revenue uplift to prospective buyers, not just time savings. Fierce Healthcare documented confirmed revenue gains at Rush, McLeod Health, and FMOL Health following Suki AI scribe deployment. The Washington Post (March 12, 2026) traced the scribe-to-billing link for a mainstream audience.
Countervailing evidence complicates the productivity justification. A 2026 study reported only “modest reductions in EHR/documentation time” (Healthcare Dive, April 2, 2026), undercutting the claim that time savings alone justify adoption. If the tools deliver modest efficiency gains but measurable revenue gains, the economic incentive tilts toward documentation volume rather than clinical efficiency.
The Attribution Problem and the Macro Cost Picture
Both parties to the dispute are self-interested. BCBSA’s study is an association-authored analysis of its own claims data, not peer-reviewed and not independently replicated. The AHA’s counter, framed in coverage as “Hospitals Say Patients Are Sicker,” is equally self-interested and equally unproven.
The 2023–2024 window is unusually confounded. Post-COVID acuity, GLP-1 drug costs, labor inflation, and the 2023 E/M coding-rule revisions all moved claims in the same period. Isolating AI coding’s contribution from those forces requires methodology neither side has publicly released in replicable form.
The $942 million figure is best read as a claim, not a finding, and as a rate-of-change signal against roughly $5 trillion in annual national health expenditure. The macro backdrop gives the claim its political force. PwC’s 9% medical cost trend projection for 2027 names AI-enabled coding alongside GLP-1 drugs. ACA marketplace premiums are proposed to rise about 14% for 2027, and the 2.48% Medicare Advantage rate increase pushes margin pressure onto both plans and providers.
The result is a competitive equilibrium. Hospitals deploy AI to document and bill more; insurers deploy AI to deny and downcode. Cigna and Aetna have pursued downcoding policies, and the CMS WISeR model covers more than 6 million Medicare beneficiaries across 17 procedure categories. Each side’s AI investment is, in part, a response to the other’s.
Industry Reaction, Regulation, and the Governance Gap
The AHA’s formal rebuttal and fact sheet have hardened the standoff. Meanwhile, counter-evidence on AI’s cost effects cuts both ways: UnitedHealth reported that AI trimmed medical costs by 270 basis points; Oliver Wyman estimated AI could erase up to 60% of the looming cost surge; and the Peterson Health Technology Institute (PHTI) found that AI raises costs for both sides depending on deployment.
Denial accuracy is under its own scrutiny. A June 2026 HHS Office of Inspector General review found that major Medicare Advantage insurers appear to deny post-acute care for profit, and patients won roughly 90% of appealed AI-driven denials.
Regulation is moving faster than the evidence. Fourteen new state healthcare AI laws passed in 2026, and North Carolina is advancing the first bill to regulate AI in medical billing. 82% of Americans back human review of AI claim denials (Medical Economics, October 29, 2026), even as the Trump administration pushes to preempt state AI insurance laws. The governance gap remains wide: as Chief Healthcare Executive (October 28, 2026) put it, “Almost all health organizations don’t know what AI they’re using.”
The Bottom Line
The $942 million figure reframes a technical billing dispute as a national premium story, but the causal question remains unresolved. Watch for BCBSA’s methodology release and any independent replication; 2027 rate filings that explicitly name AI coding; the North Carolina Senate billing bill; CMS WISeR evaluation data; court-ordered discovery in the UnitedHealth nH Predict litigation; and whether the October 26 “affordability pledge” produces anything enforceable.
The transferable insight is a Jevons paradox for administrative labor: AI lowers the cost of documentation, which expands the volume of billable events rather than shrinking total spend. Until someone measures that effect independently, both the $942 million claim and the hospitals’ rebuttal remain assertions in an argument that premiums will settle.