AI in Healthcare is Advancing. Is Healthcare Getting Better?

11 September, 2026

By Gold Igwe

The debate over artificial intelligence has become increasingly tied to the United States’ broader competition for technological and economic leadership. President Donald Trump has recently pushed back against warnings that AI could lead to human extinction while emphasizing the importance of keeping the United States ahead of China in the development of the technology, according to Bloomberg. In healthcare, however, the consequences are already being determined at a much more practical level, as physicians and healthcare organizations decide which applications are useful enough to incorporate into everyday work and which require greater evidence, oversight, or limits before they can be trusted.

Doximity’s 2026 State of AI in Medicine report, based on responses from 3,151 U.S. physicians across 15 specialties, found that 94% were either using AI or interested in using it, while 54% reported actually using AI in their practice. Reported use increased from 47% in the April 2025 cohort to 63% in the January 2026 cohort. The American Medical Association’s 2026 physician survey found that 81% of physicians now use AI professionally, compared with 38% in 2023, and more than three-quarters said the technology improves their ability to care for patients.

The applications attracting the most use are also among the easiest to evaluate within an existing clinical workflow. Doximity identified literature search and voice-based documentation among the leading uses, while the AMA found that physicians most commonly use AI to summarize medical research and standards of care, create discharge instructions and care plans, document billing codes and medical visits, and generate chart summaries. Assistive diagnosis, by comparison, was reported by 17% of physicians in the AMA survey.

Physicians are also becoming more demanding about the conditions under which they use these systems. Accuracy and reliability were the leading concerns in Doximity’s survey, cited by 71% of respondents, while the AMA found that 86% consider data privacy important to broader adoption and 88% want robust safety and efficacy validation. Eighty-five percent said they want to be consulted or directly involved in decisions about AI adoption, and physicians also identified clear liability frameworks as an important part of establishing trust in the technology.

A generated clinical note can be reviewed against the encounter before it is used, while an AI system that influences diagnosis or treatment introduces a much higher degree of responsibility for the physician and the organization deploying it. That difference helps explain why administrative documentation, research, information retrieval and other assistive applications have gained traction faster than systems designed to take a more direct role in clinical judgment.

The technology is now moving further into the systems that support patient care. On September 1, OpenAI announced an integration between ChatGPT Health and Epic’s electronic health record system, which holds information on more than 325 million patients. The integration allows clinicians to bring appointment notes, laboratory results, medications, specialist documentation, and other patient information into ChatGPT for tasks including summarizing medical histories, identifying changes and preparing for appointments. The connection is read-only and does not allow the AI system to write information back into the health record. TechCrunch

As AI becomes more closely integrated with healthcare systems, its potential value grows alongside the consequences of getting implementation wrong. A product that reduces documentation time can increase the capacity available to physicians and reduce administrative workload, whereas unreliable outputs, mishandled patient information, poor workflow design, or inadequate security can create costs involving privacy, cybersecurity, regulatory compliance, liability, and patient safety. A healthcare AI company, therefore, has to demonstrate more than technical capability if it expects its product to become embedded in a provider’s operations; the technology has to function reliably within the environment where people, data, regulations, and clinical responsibilities intersect.

There are already large pools of healthcare work where that economic case can be tested. Documentation, information retrieval, workflow coordination, administrative processing, and other repetitive activities consume professional time that healthcare organizations have an obvious financial reason to recover. Products that can reduce those costs, improve capacity, or make existing processes more efficient have a concrete starting point for demonstrating value, particularly when they can integrate with software that providers already use.

Products that move closer to autonomous clinical decision-making face a different set of requirements. Evidence of accuracy, appropriate validation, privacy protections, explainability, physician oversight, and clearly defined responsibility become increasingly important as the consequences of an error become harder to contain. The AMA has emphasized many of these considerations in its broader guidance on augmented intelligence in medicine, including the need for clinical evidence, responsible implementation, transparency, and physician involvement.

Meanwhile, access to increasingly capable AI models is becoming less exclusive, which may make the model itself a weaker source of long-term differentiation. Businesses that have built strong integrations with healthcare systems, accumulated relevant data, demonstrated performance in a particular workflow, developed regulatory and security capabilities, or established trusted relationships with healthcare providers have advantages that are considerably harder to reproduce than access to a general-purpose model.

The U.S. healthcare AI market is consequently developing as adoption rises alongside scrutiny. Physicians are using AI more frequently and reporting benefits from applications that reduce administrative work and improve access to information, while concerns about accuracy, privacy, safety, liability, and oversight remain significant. Businesses operating in the sector will have to address both sides of that equation if they are to move beyond experimentation and become part of the systems healthcare organizations rely on every day.

Some of the most compelling opportunities may therefore lie in businesses that use AI to solve expensive, persistent problems without requiring healthcare organizations to surrender the judgment and accountability that remain central to patient care. Reducing administrative costs, improving professional capacity, strengthening existing workflows, and earning the confidence of physicians can create substantial economic value even when AI remains an assistive technology. As these systems become increasingly embedded in U.S. healthcare, the businesses that can translate that value into durable operations, defensible capabilities, and responsible implementation will have the strongest foundation for lasting growth.