How to use artificial intelligence (AI) and large language models (LLMs) to understand your health information

An increasing number of U.S. adults say they have used an AI chatbot to learn more about their health.

Here are our top tips and advice for using AI when learning more about your own health information.

Artificial intelligence, or AI, is a type of technology. It learns from large amounts of information and solves certain problems. You may have come across AI in everyday life. It helps show traffic updates on smartphones. It can suggest products to you based on your online shopping history. It can flag unusual activity on bank accounts.

AI is being adopted in healthcare. It may help doctors diagnose diseases. It can support medical research. It can help catch possible medical errors.

Some doctors are using AI to take notes during patient visits. Others are using it to search for recent medical studies. When used by a health care organization, these tools are designed to follow privacy rules.
AI shows promise, but it is a developing technology. Researchers and doctors continue to study how well these tools work. They are looking at how to use them safely and responsibly.

A large language model (LLM) is a type of artificial intelligence. It is a computer program that learned human language by reading large amounts of text. LLMs look for patterns in language and use those patterns to have conversations, answer questions, or predict the next word used in a sentence. You may already have heard of some of these tools such as ChatGPT, Claude, or Gemini.

LLMs feel different from older computer programs. They seem more chatty, friendly, and “real” because they are based on our human language.

There are very few studies on this topic. One study tested a possible approach to using AI to understand an open note. It found that people could get more accurate health information when they asked the AI to respond as if it were a doctor. This study was “proof-of-concept,” which means it was an early test of an idea. The researchers believe more work is needed to test this idea.

After a medical visit, doctors, nurses, and other health care providers write notes that summarize important information about you. These notes become a part of your medical record. When a note is shared with you, it becomes an “open note.”

An open note may include:

  • A summary of what you told the doctor or nurse, also called a “history.”
  • The provider’s findings from a physical exam. Examples may include: blood pressure, weight, how your lungs or heart sound, a description of a growth on your skin, or observations about your mood.
  • Your provider’s thoughts about results of your lab tests, x-rays, scans, biopsies, or other tests.
  • Summary thoughts about any medical conditions or symptoms. This section may also be called “assessment” or “impressions.”
  • Recommendations made during the visit. This section is often called the “treatment plan” or “plan of care.” Notes might include ordered tests, prescribed medications, follow-up appointments, referrals, exercises, or recommended changes to your diet.

How notes look depend on many things, including: who wrote the note, the kind of visit you had, or where you get care. Some notes are short. Others might include a complete description of your visit. Longer notes may include additional details about your health, such as past problems or test results. Some notes may include information that doesn’t seem relevant due to various insurance requirements or other rules and regulations.

The most common way to access your notes is through the online secure patient portal which is hosted by your healthcare provider. Using this secure and free portal, you can, among other things, request appointments, message your health care team, and view your health record, including the notes. If your doctor or nurse does not use a web portal, you can request a digital or paper copy of these records for free. If you are not registered on the portal, contact your provider’s office to find out how to create a portal account.

Some people use AI chatbots to better understand their health. One approach is to ask the chatbot to “interview me as if you’re a doctor.” This approach may help you prepare for doctor visits. However, this is not a replacement for medical advice. Experts at Beth Israel Deaconess Medical Center suggest the following steps:

  1. Pull up your medical notes.
  2. Remove any personal information that can identify you.
  3. Copy the notes into an AI tool.
  4. Give the AI a current update on your health concerns.
  5. Ask the chatbot to summarize the information.
  6. Then ask: “Given this context about my health, give me three questions I should ask my doctor about my condition during my upcoming visit.”

Large language models (LLMs) want to please their users and were designed for “engagement”. They pick up on information that resonates with you. LLMs do not have built-in fact checkers and may give you inaccurate information. This is troublesome for people using AI to answer health questions. It will expose you to more of that information because it is assuming that is what you want.

Experts are advising patients to tell AI why they are asking questions. Instead of saying ‘I have a headache. What should I do?’ try saying ‘I am having a bad headache today. Here is my last note from my primary care doctor. What are some strategies to make it better?’

Experts encourage patients to think carefully before sharing personal health information with AI tools. Ask yourself who you trust with your information and what might happen to it over time. Once you share information with an AI platform, the company that runs the tool may store or use that information.

It is important to remove details that could identify you. This includes your medical record number, Social Security number, phone number, and home address.

Bias in health care happens when people are treated unfairly. This happens because of people’s attitudes and beliefs about, for example, race, age, gender, language, body size, disability, or income, among other things. These biases can affect the care people receive and lead to worse health outcomes for some groups. In the past, health information has been used to discriminate against people with illnesses or disabilities. This kind of bias can affect artificial intelligence (AI).

AI is now used to help make decisions in health care. For example, AI tools may help decide who gets certain treatments, health insurance approval, or who is seen as high risk for illness. Most patients cannot see these tools, but they may still affect the care they receive.

Insurance companies may use AI to predict health risks or to review claims. Some AI tools are used to flag patient claims for denial. Health care systems may use AI systems that are trained on biased data. When AI is trained on data that reflects existing inequalities, it can worsen health disparities.

Many AI companies are for-profit businesses that work to make money for shareholders. While doctors must follow strict rules because of their medical licenses, companies that build AI tools may not have to follow the same rules.

Because of this, it is important for you to think carefully about whether health AI tools are working in your best interest.

OpenNotes is a research program that studies how sharing health information affects patients and clinicians. Our work is supported by grants from foundations, government agencies, and technology companies. These funders do not control our research methods or findings. We share what we learn to help patients and care teams make informed decisions.