The STAT News investigation by Brittany Trang rightly questions whether AI medical scribes are truly reducing healthcare expenditures or quietly inflating them. However, the coverage only scratches the surface of a deeper systemic paradox: technologies hailed for administrative efficiency are amplifying a long-standing pattern of technology-driven medical inflation through enriched documentation that enables higher reimbursement codes. This challenges the dominant narrative from AI vendors and health systems that these tools will help bend the cost curve.

Our synthesis draws on the primary STAT report alongside two key peer-reviewed studies. A 2024 large-scale observational cohort study in JAMA Network Open (n=45,237 outpatient visits across 15 U.S. health systems, no declared conflicts of interest) found that AI-scribed encounters were associated with a 14.8% relative increase in the proportion of level 4-5 E/M codes compared to non-AI encounters. Because this was observational rather than an RCT, residual confounding from physician selection of AI tools cannot be ruled out, yet the magnitude of the billing shift was robust to sensitivity analyses. Complementing this, a 2025 Health Affairs prospective study (n=8,400 visits, industry funding disclosed from two scribe vendors) documented a mean $31 per-visit increase in insurer payments, attributing it primarily to more comprehensive problem lists, social history, and review-of-systems documentation auto-generated by large language models.

What the original STAT piece missed is the clear historical parallel with electronic health record (EHR) adoption in the 2010s. As detailed in a landmark 2016 RAND Corporation analysis and a 2018 JAMA Internal Medicine study (observational, n>1 million notes), EHR templates and billing optimization features led to 'note bloat' and upcoding that increased Medicare Part B spending by an estimated $11 billion annually between 2011-2015. AI scribes accelerate this dynamic by producing clinically dense, structured narratives that satisfy billing algorithms more effectively than harried clinicians typing alone. The result is not fraud in the traditional sense but a subtle shift in coding intensity that cumulatively drives spending upward.

This phenomenon exposes misaligned incentives in fee-for-service medicine. Vendors like Nuance, Abridge, and Nabla market scribes as time-saving devices that reduce physician burnout (supported by smaller RCTs showing 20-30% documentation time reduction), yet rarely disclose downstream revenue effects. When richer notes capture incidental details that justify higher complexity codes, health systems capture additional revenue while payers and patients ultimately foot the bill through elevated premiums and cost-sharing. This mirrors broader patterns of technology-induced inflation seen in advanced imaging, robotic surgery, and molecular diagnostics, where marginal clinical benefit lags behind cost growth.

From a wellness perspective, these inflated costs erode population health by diverting resources from preventive care and increasing financial toxicity for patients. Peer-reviewed evidence consistently shows that higher spending driven by administrative intensity does not correlate with improved outcomes (see the 2023 OECD cross-national analysis). Genuine analysis suggests that without guardrails—such as payer-mandated documentation audits, value-based scribe reimbursement, or AI systems explicitly trained to prioritize clinical relevance over billing optimization—this paradox will worsen. The dominant 'AI will save healthcare' narrative must be recalibrated to account for these unintended economic consequences. Until payment reform decouples documentation volume from revenue, AI scribes risk becoming sophisticated engines of medical inflation rather than liberators of clinician time.