The essentials

An "AI doctor" is a chatbot or app that answers health questions, sorts symptoms or explains test results. It helps you understand words and numbers and prepare questions. It cannot examine you, it can be confidently wrong, and in studies people using chatbots did no better at recognizing their condition. Use it to understand, never to decide alone.

What can an AI doctor actually do?

"AI doctor" is not a medical title but a label for several tools you can use alone, at any hour. In 2024, 17% of US adults said they use AI chatbots at least once a month for health information, and 25% of adults under 30[1].

  • Symptom checkers: you answer questions and the app suggests possible causes and where to go. They existed long before chatbots[2].
  • General chatbots: assistants that answer health questions in everyday language, like any other question. They were not designed as medical tools.
  • Result explainers: services that put each lab value in plain words next to its reference range, as described in our guide to AI lab interpretation.
  • Tools for clinicians: software that supports doctors' decisions, some of it regulated as medical device software in the United States[8].

Their strength is language: explaining a term, rephrasing a doctor's note, listing questions. None of them can listen to your chest, see how you look, or know what you did not type. For the wider picture, see the benefits of AI in healthcare.

How accurate are AI doctors? What the studies show

Chatbots do very well on written cases. Results are much weaker when real people describe their own problem.

Published studies on AI for symptoms and medical questions
StudyWhat was testedMain result
BMJ, 2015[2]23 symptom checkers, 45 written casesCorrect diagnosis first in 34%; right advice on where to go in 57%
Review, 2023[3]14 studies of symptom checkersDiagnostic accuracy low, almost always below doctors or nurses; triage below par in 9 of 13 studies
JAMA Intern Med, 2023[4]195 forum questions answered by doctors and a chatbotProfessionals preferred the chatbot answer in 78.6% of ratings
Trial, 2024[5]50 doctors on clinical cases, with or without a chatbot76% with it, 74% without (no real difference); the chatbot alone scored 16 points higher
Trial, 2026[6]1,298 UK adults, 10 scenarios, with a chatbot or their usual sourcesChatbots alone found a relevant condition in 94.9%; people using them in fewer than 34.5%, less often than people using search engines or health websites

The last study matters most for patients. People often gave only part of the story (16 of 30 conversations reviewed), the chatbot sometimes misread the question, and people did not always pick the right suggestion. They chose the right course of action in fewer than 44.2% of cases, no better than the comparison group, and exam-style tests did not predict these failures[6].

The forum study judged how good answers looked, not whether patients got better, and doctors' replies were much shorter (52 words against 211)[4]. Good writing is not a correct diagnosis.

What are the risks of asking AI about your symptoms?

  • Common Missing context. A chatbot does not see that you are pale or confused, and may not ask about your medicines, pregnancy or past illnesses.
  • Common Confident errors. The World Health Organization warns of false, inaccurate, biased or incomplete statements[7]. A wrong answer reads like a right one.
  • Possible Over-trust. WHO also describes "automation bias": accepting the machine's answer and missing its mistakes[7].
  • Possible Wrong urgency. In the 2026 study, people tended to underestimate urgency, with or without a chatbot, and two people describing very similar signs of a brain bleed got opposite advice: rest in a dark room, or emergency care[6].
  • Possible Privacy. What you type may be stored or used to improve the service, and many consumer apps fall outside HIPAA[9].

With lab results, one value flagged H or L often means little, as explained in abnormal results when you feel fine, while a few need fast action (lab errors vs real emergencies). A chatbot does not always tell them apart.

When not to ask a chatbot first

  • Emergency: call 911 for chest pain, signs of a stroke (face drooping, weak arm, trouble speaking), severe trouble breathing, fainting, heavy bleeding, confusion, or fever with feeling very unwell, which can be sepsis.
  • Within 24 to 48 hours if symptoms keep getting worse, a lab or doctor has called you about a result, or an app advised you to be seen quickly.
  • At your next appointment for the rest, including what an AI tool told you.

Who regulates AI doctors and your health data?

  • FDA (United States): software meant to diagnose or treat can be a medical device. The FDA's January 2026 guidance on clinical decision support explains which tools for professionals are excluded, and its digital health policies still apply to device software for patients[8]. If an app does not say it is FDA-authorized for a medical use, assume it was not reviewed as a device.
  • Privacy (United States): HIPAA covers your doctors, hospitals and insurers. Companies outside HIPAA still answer to the FTC Act, and some must report breaches under the Health Breach Notification Rule[9].
  • GDPR (European Union): health data is sensitive data; processing it is prohibited except in specific cases, such as explicit consent[11].
  • AI Act (European Union): since August 2026, people must know when they are interacting with a machine. Rules for high-risk AI in regulated products, such as medical device software, apply from August 2, 2028[10].
  • Health data hosting: in France, hosting personal health data is regulated and hosts can be certified "HDS"[12].

How can you use an AI doctor safely?

  • Use it to understand, not to decide.
  • Give context: age, pregnancy, conditions, medicines, how long symptoms have lasted.
  • Ask what would make your situation urgent, and act on red flags without waiting.
  • Remove your name, birth date and insurance numbers before sharing a document; read how data is stored.
  • Never start, stop or change a medicine because a chatbot suggested it.
  • Use the reference range on your own report, as explained in how to read blood test results.

Questions to ask your doctor after using an AI tool

  1. The app suggested this cause: does it fit my exam and history?
  2. Which of my results really matter?
  3. Is anything I read a reason to come back sooner?

What does BloodSense do, and what doesn't it do?

  • What it does: you upload a blood, urine or stool lab report you already have. It explains each line against the reference range on your report, reads related results together (such as a complete blood count or a metabolic panel), and lists questions for your doctor. See the blood test interpretation page.
  • What it does not do: examine you, diagnose, prescribe or replace a visit. It works from your report, cannot know what you have not shared, and is not for emergencies.
  • Your data: your name is removed before analysis, and health data is stored on HDS-certified servers in France (see the FAQ and the privacy policy).

Have a full lab report in front of you? BloodSense explains every line in plain words, shows how results relate, and gives you clear questions for your doctor.

Explain my lab report

Frequently asked questions

Can AI replace my doctor?

No. A doctor examines you, knows your history, orders tests, prescribes and takes responsibility. In a 2024 trial, giving doctors a chatbot did not improve their diagnostic reasoning[5].

Is ChatGPT accurate for medical advice?

Often for general information, but it can be wrong convincingly, and accuracy drops when people describe symptoms in their own words[6]. Treat answers as something to check.

Why does the AI answer differently when I ask again?

Chatbots compose each answer anew and react to small changes in wording. A different answer is a reason to check with a professional, not to pick the one you prefer.

Are symptom checker apps better than chatbots?

Not clearly: symptom checkers had low diagnostic accuracy[3], and chatbots have not done better with the public[6].

  1. KFF. KFF Health Misinformation Tracking Poll: artificial intelligence and health information. August 15, 2024. kff.org (accessed November 3, 2026)
  2. Semigran HL, Linder JA, Gidengil C, Mehrotra A. Evaluation of symptom checkers for self diagnosis and triage: audit study. BMJ. 2015;351:h3480. doi:10.1136/bmj.h3480 (accessed November 3, 2026)
  3. Riboli-Sasco E, El-Osta A, Alaa A, et al. Triage and diagnostic accuracy of online symptom checkers: systematic review. J Med Internet Res. 2023;25:e43803. doi:10.2196/43803 (accessed November 3, 2026)
  4. Ayers JW, Poliak A, Dredze M, et al. Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum. JAMA Intern Med. 2023;183(6):589-596. doi:10.1001/jamainternmed.2023.1838 (accessed November 3, 2026)
  5. Goh E, Gallo R, Hom J, et al. Large language model influence on diagnostic reasoning: a randomized clinical trial. JAMA Netw Open. 2024;7(10):e2440969. doi:10.1001/jamanetworkopen.2024.40969 (accessed November 3, 2026)
  6. Bean AM, Payne RE, Parsons G, et al. Reliability of LLMs as medical assistants for the general public: a randomized preregistered study. Nat Med. 2026;32(2):609-615. doi:10.1038/s41591-025-04074-y (accessed November 3, 2026)
  7. World Health Organization. WHO releases AI ethics and governance guidance for large multi-modal models. January 18, 2024. who.int (accessed November 3, 2026)
  8. US Food and Drug Administration. Clinical decision support software: guidance for industry and FDA staff. January 2026. fda.gov (accessed November 3, 2026)
  9. Federal Trade Commission. Collecting, using, or sharing consumer health information? Look to HIPAA, the FTC Act, and the Health Breach Notification Rule. August 2024. ftc.gov (accessed November 3, 2026)
  10. European Commission. AI Act: regulatory framework on artificial intelligence (Regulation (EU) 2024/1689), application timeline. digital-strategy.ec.europa.eu (accessed November 3, 2026)
  11. European Data Protection Board. Data protection basics (SME data protection guide): sensitive data. edpb.europa.eu (accessed November 3, 2026)
  12. Agence du numérique en santé. Hébergement de données de santé (HDS). Updated September 22, 2026. esante.gouv.fr (accessed November 3, 2026)

This page is general information about AI tools in health. It does not replace the advice of a doctor, who examines you, interprets your results in context and decides on tests and treatment.