The AI scribe story just got a lot more complicated. A new report from the Blue Cross Blue Shield Association found that AI-driven billing practices have contributed to a $2.3 billion increase in U.S. healthcare spending — driven primarily by a surge in upcoding, where AI tools capture clinical encounters so thoroughly that visits get billed at higher-acuity levels than they might have been before. The mechanism is almost straightforward: burned-out doctors used to take shortcuts on paperwork. AI scribes don't. More complete notes support more complex billing codes — and the math adds up fast.

$2.3B

Increase in U.S. healthcare spending attributed to AI-driven billing practices since 2024, per Blue Cross Blue Shield Association — with $663M in inpatient and $1.67B in outpatient costs.

Behind closed doors, the industry's two loudest adversaries — hospitals and insurers — are in rare agreement. At a recent Peterson Health Technology Institute roundtable, every stakeholder in the room acknowledged the problem privately. Insurers are now deploying their own AI to "downcode" visits and claw back payments, setting up what health economists are calling an "AI billing arms race" — one that smaller and rural providers, who haven't adopted scribes yet, may end up losing even though they had nothing to do with starting it.

The irony is sharp. The technology most celebrated for reducing physician burnout may be accelerating the same affordability crisis that burnout was supposed to help solve. The industry needs a reckoning on how AI-assisted documentation interacts with billing incentives — before regulators force one.

A sweeping new poll from the West Health-Gallup Center on Healthcare in America, published April 15, captures something that's been happening quietly in waiting rooms across the country: patients are increasingly arriving — or not arriving — having already consulted an AI. Twenty-five percent of U.S. adults have now used an AI tool or chatbot for health information or advice, with the majority using it to research before or after seeing a doctor.

14M

Adults who skipped a provider visit because AI-generated health information led them to believe they didn't need one — representing 14% of recent AI health users, per the West Health-Gallup national poll.

The motivations are revealing. Some patients use AI simply for convenience or curiosity. But a meaningful share turn to it out of necessity: 14% because they couldn't afford a doctor visit, 16% because they couldn't access a provider, and 21% because a provider had dismissed them in the past. For this group, AI isn't a tech novelty — it's filling a gap the healthcare system created. About half said AI made them feel more confident when eventually speaking with a clinician, and 22% said it helped them identify issues earlier. Yet trust remains thin: only 4% of users strongly trust the accuracy of AI-generated health information, meaning millions of Americans are making meaningful healthcare decisions on information they're not sure they can believe.

03 / Regulation

On March 20, the White House released a National Policy Framework for Artificial Intelligence, and its implications for healthcare are significant enough that legal and compliance teams should not be treating this as background noise. The framework calls on Congress to establish a federal standard that would preempt state AI laws imposing "undue burdens" — a direct challenge to the growing patchwork of state-level AI healthcare rules that organizations are currently trying to navigate simultaneously.

The critical caveat: no legislation has passed yet. State laws — including the Utah AI Act, the Colorado AI Act, and California's AI privacy amendments — remain fully in effect. The Department of Commerce was asked to publish an evaluation of existing state AI laws by March 11; that report has not yet been released. In the meantime, states are continuing to pass new laws, and any eventual federal preemption will almost certainly face legal challenges.

The framework also flags children and minors as a priority focus, which has direct implications for pediatric practices, children's hospitals, and any AI tool — including mental health chatbots — that might interact with patients under 18. Holland & Knight this week launched a dedicated Health AI Navigator, an interactive tracker of federal and state AI legislative activity, as a practical response to the compliance complexity that's now routine in healthcare legal departments. The message is clear: the regulatory moment is not coming — it is here, and it is moving in multiple directions at once.

Sources