For years, artificial intelligence in health care was largely discussed as a future technology. Doctors might someday use AI to identify diseases, analyze medical images, or help determine which treatments could work best.
That future is arriving much faster than many patients may realize.
Health systems are increasingly using artificial intelligence to search through and summarize electronic health records, potentially allowing doctors to find important information buried deep inside years of medical notes, test results, and diagnoses. A report published this week by STAT described several health systems moving toward broader use of AI chatbots designed specifically to query patient records.
The development could make medicine more efficient and help doctors uncover information they might otherwise miss. But it also raises a difficult question that deserves more public attention.
How much should we trust an AI system with our medical history?
America’s Medical Records Are Already Digital
The rise of AI in medical records would not be possible without another transformation that has already taken place.
Americans have spent years moving from paper medical charts to electronic health records. According to the Office of the National Coordinator for Health Information Technology, 91% of office-based physicians and more than 99% of non-federal acute-care hospitals had adopted certified electronic health records by 2024.
That means enormous amounts of information about patients are now stored digitally.
A medical record can contain years of information. A patient might have laboratory results from multiple facilities, imaging reports, prescriptions, diagnoses, specialist consultations, emergency room visits, and physician notes. Some of the most important information may be buried in a record created years earlier.
Finding that information can be difficult even for experienced physicians.
That is where AI can become useful.
Finding the Needle in the Haystack
That same STAT report detailed a case at Stanford involving a patient whose doctors were struggling to identify the cause of cancer found in a lymph node. A physician used an AI tool called ChatEHR to search the patient’s history and discovered that the patient had previously been diagnosed with a related cancer at another health system. That information helped explain the new findings.
The example illustrates why AI could become an important tool in medicine.
An AI system does not necessarily need to make the diagnosis itself to be useful. It can instead help a physician locate information that already exists.
A doctor might ask an AI system whether a patient has previously experienced a particular symptom, undergone a particular procedure, or received a particular diagnosis. Instead of manually searching through hundreds of pages of records, the physician could receive a summarized answer in seconds.
The technology could effectively function as a research assistant for the patient’s own medical history.
That may be particularly valuable for people who see numerous specialists or receive care from multiple health systems.
AI in Health Care Is Already More Common Than Many Patients Realize
The use of AI in medical records is not an isolated experiment.
Federal data show that 71% of U.S. hospitals reported using predictive AI integrated with their electronic health records in 2024, up from 66% in 2023. The technology is being used for purposes including predicting health risks, identifying high-risk patients, and supporting treatment decisions.
Hospitals are also using AI for less glamorous but potentially important tasks. The federal data show substantial growth in AI applications involving billing and scheduling.
That matters because it demonstrates that artificial intelligence is becoming embedded in the health care system even when patients may not realize it.
The AI revolution in medicine is therefore not simply about robots performing surgery or chatbots diagnosing diseases.
Increasingly, it is about software operating behind the scenes, processing information contained in the electronic records that physicians already use.
Doctors Are Becoming More Comfortable With AI
Physicians themselves appear to be increasingly accepting of the technology.
The American Medical Association’s 2026 physician survey found that awareness and use of AI in medical practice continued to grow. The organization reported that more than 80% of physicians now use AI professionally or have used it in some capacity.
That does not mean doctors are handing control of patient care over to machines.
In many cases, AI is being used to reduce administrative work, summarize documentation, organize information, or provide another tool for physicians to consider.
That distinction will become increasingly important as AI becomes more sophisticated.
The safest role for artificial intelligence may not be replacing the physician. It may be helping the physician make better use of the information already available.
The Problem With an AI Summary
There is, however, an obvious danger.
AI systems can make mistakes.
A system that summarizes a medical record might misunderstand a doctor’s note, overlook an important detail, or present information without sufficient context. A generative AI system could potentially produce an answer that sounds confident even when the underlying information is incomplete.
That is particularly concerning in medicine because the consequences of an error can be far more serious than getting a wrong answer from an ordinary chatbot.
A physician who misses a minor detail in a travel recommendation can correct the mistake later. A physician who overlooks an important medical history item could make a decision that affects someone’s health.
This is why federal research into hospital AI adoption is encouraging in one respect. The Office of the National Coordinator found that 82% of hospitals using predictive AI reported evaluating models for accuracy, 74% evaluated them for bias, and 79% conducted post-implementation evaluation or monitoring in 2024.
Hospitals are also using AI for less glamorous but potentially important tasks. The federal data show substantial growth in AI applications involving billing and scheduling.
This demonstrates that artificial intelligence is becoming embedded in the health care system even when patients may not realize it.
The AI revolution in medicine is therefore not simply about robots performing surgery or chatbots diagnosing diseases.
Increasingly, it is about software operating behind the scenes, processing information contained in the electronic records that physicians already use.
Those numbers are encouraging, but they also demonstrate that AI requires continuing oversight.
A model that performs well today cannot simply be installed and forgotten.
Your Medical Records Are Different From Other Data
There is another reason this technology deserves scrutiny.
Medical records are among the most sensitive forms of personal information people possess.
They can contain information about illnesses, medications, mental health, family history, reproductive health, substance use and other deeply personal matters. A medical record can tell a story about someone that they might not even share with close friends or family members.
Giving AI access to that information therefore creates privacy and security questions that do not exist to the same degree when someone asks an AI system to write an email.
The Department of Health and Human Services has emphasized the importance of privacy and security when health organizations adopt AI. Its health care cybersecurity guidance specifically warns organizations to establish protocols for assessing data privacy and security risks when selecting and implementing AI technologies.
The question is not simply whether an AI system can understand a medical record.
It is also who can access the system, where the information is stored, how it is protected and what happens to the information after the AI has processed it.
The Data Problem Could Become Even Bigger
The amount of medical information available to AI systems is likely to grow.
Federal health IT officials are actively working on standards intended to make electronic health information easier to exchange between systems. The Office of the National Coordinator released the seventh version of the United States Core Data for Interoperability in July, continuing efforts to standardize the information that can be exchanged between health systems.
Better interoperability could be enormously beneficial.
A patient’s medical history should ideally follow the patient, rather than becoming fragmented among hospitals, specialists, and doctors’ offices.
But greater accessibility also creates greater responsibility.
The more information that becomes available to AI systems, the more important it becomes to establish clear rules governing how that information is used.
The Digital Divide Could Affect AI in Medicine
There is another issue that could receive less attention than privacy or accuracy.
Not every hospital has the same resources.
Federal research found that AI adoption was considerably lower among small, rural, independent, and critical-access hospitals than among larger, urban, and system-affiliated hospitals. In 2024, 56% of rural hospitals reported using predictive AI compared with 81% of urban hospitals.
That raises an important question about whether AI will improve health care equally for everyone.
A large hospital system may have the money and technical staff to evaluate sophisticated AI tools. A small rural hospital may not.
If AI becomes an increasingly important part of medical care, access to the technology could eventually become another factor separating well-resourced health systems from those struggling to keep up.
AI Should Help Doctors, Not Replace Them
There is an understandable temptation to portray artificial intelligence as either the solution to America’s health care problems or an inevitable threat to patients.
Neither description is particularly useful.
The technology is a tool.
Used properly, AI could help physicians navigate enormous medical records, identify information that might otherwise be overlooked, and spend less time on administrative tasks. It could potentially give doctors more time to talk with patients instead of searching through computer screens.
But those benefits depend on maintaining human oversight.
A doctor should be able to question an AI-generated summary, examine the underlying records, and ultimately make an independent clinical judgment. Patients should also know when AI is being used in their care and understand what role it plays.
The Medical Records of the Future
The most significant change may be that the medical record itself is becoming something more than a digital filing cabinet.
It is increasingly becoming a source of information that machines can search, organize, and interpret.
That could fundamentally change how physicians interact with patient histories.
Instead of spending valuable time looking for a diagnosis buried three years earlier in another hospital’s records, a doctor could ask a question and receive the relevant information almost immediately.
That is an exciting possibility.
It is also a responsibility that should not be taken lightly.
The medical records of millions of Americans contain some of the most private information imaginable. As artificial intelligence gains greater access to those records, patients need confidence that the technology is accurate, secure, and being used responsibly.
The future of AI in medicine may ultimately depend less on whether machines can read our medical records than on whether Americans can trust the people and institutions allowing them to do it.
—Greg Collier
