Several free online tools let you enter your symptoms and receive a list of possible diagnoses, ranked by likelihood. The most widely used options include Mayo Clinic’s Symptom Checker, WebMD’s Symptom Checker, and Ada Health, an AI-powered app. None of them replace a doctor’s evaluation, and their accuracy varies more than you might expect, but they can be a reasonable starting point for understanding what might be going on.
The Main Symptom Checker Tools
Most symptom checkers follow the same basic pattern: you select or type your symptoms, answer follow-up questions about timing, severity, and your medical history, and then receive a ranked list of conditions that could explain what you’re experiencing. Here are the major options:
- Mayo Clinic Symptom Checker (free, web-based): You pick a symptom from a list and answer guided questions. It’s straightforward and backed by Mayo’s medical database, though it tends to cast a wide net.
- WebMD Symptom Checker (free, web and app): One of the oldest tools available. You select a body area, then check off symptoms. It returns a list of possible conditions with links to more detailed reading.
- Ada Health (free, app-based): Uses an AI-driven conversation format. It asks detailed follow-up questions that narrow down possibilities more aggressively than most competitors. It’s generally considered one of the more thorough dedicated symptom checkers.
- ChatGPT and other AI chatbots: You can describe your symptoms in plain language and get a conversational response. These aren’t purpose-built medical tools, but they’ve been tested head-to-head against dedicated symptom checkers with mixed results.
How Accurate These Tools Actually Are
This is where expectations need adjusting. A systematic review of symptom checker studies found that the correct diagnosis appeared as the tool’s top suggestion only 19 to 38% of the time. That number improves when you look at whether the right answer appears anywhere in the top three suggestions, but it’s still far from reliable on its own.
A study comparing Ada Health, WebMD, ChatGPT 3.5, and ChatGPT 4.0 against emergency department physicians found that physicians got the top diagnosis right 47% of the time. The best-performing digital tools matched at about 40%. When expanded to the top three suggestions, physicians reached 69% accuracy, while the tools ranged from 50 to 63%. So the gap between a doctor and an app is real, but perhaps smaller than you’d assume.
One interesting finding from rheumatology research: Ada Health actually outperformed physicians in correctly classifying inflammatory rheumatic diseases, getting it right 70% of the time compared to 54% for doctors. That result likely reflects how rare and tricky those conditions are to spot in a brief appointment. Symptom checkers don’t get tired, distracted, or anchored to their first impression the way humans sometimes do.
How These Tools Decide What’s Wrong
Most dedicated symptom checkers use a type of statistical model called a Bayesian network. In simple terms, the system knows that certain symptoms tend to appear together in specific conditions. When you report a combination of symptoms, it calculates the probability of various diagnoses based on how commonly those symptoms cluster in real patient data. Each follow-up question you answer shifts those probabilities, narrowing down the list.
AI chatbots like ChatGPT work differently. They’re trained on enormous amounts of text, including medical literature, and generate responses based on language patterns rather than structured medical probability models. This gives them flexibility in understanding how you describe your symptoms in everyday language, but it also means they can sometimes sound confident about wrong answers.
The Triage Problem
Getting a list of possible conditions is one thing. The more important question is often: do I need to go to the ER, see a doctor this week, or just wait it out? Symptom checkers typically sort their recommendations into urgency levels like “emergency care now,” “see a doctor within a few days,” or “safe to manage at home.”
In the emergency department comparison study, ChatGPT 4.0 agreed with physician triage decisions 76% of the time, and WebMD agreed 70% of the time. Ada Health matched at 62%. But the critical number is the “unsafe” triage rate, meaning the tool told someone their situation wasn’t urgent when it actually was. ChatGPT 3.5 had an alarming 41% unsafe triage rate, meaning it missed the seriousness of the situation in nearly half of cases that needed urgent care. Ada’s unsafe rate was much lower at 14%.
The takeaway: dedicated symptom checkers like Ada tend to err on the side of caution, sometimes sending you to a doctor when you didn’t need to go. General AI chatbots are more likely to err in the dangerous direction, reassuring you when you should be worried. If you’re using these tools to decide whether something is an emergency, a purpose-built symptom checker is the safer bet.
What Your Data Happens After You Enter It
When you type your symptoms into an app, you’re handing over sensitive health information. Most standalone symptom checkers are not covered by HIPAA, the federal law that protects your medical records at hospitals and doctor’s offices. According to the U.S. Department of Health and Human Services, once your health information is received by an app that isn’t a covered entity or business associate under HIPAA, that information is no longer subject to HIPAA protections.
In practical terms, this means the app’s own privacy policy is the only thing governing what happens to your data. Some apps share anonymized data with researchers. Others may use your information for targeted advertising. Before entering detailed symptoms, it’s worth checking whether the app sells data to third parties and whether it stores your information or deletes it after your session.
Getting the Most Useful Results
A few things can improve what you get out of a symptom checker. Be specific about timing: when did the symptom start, is it constant or intermittent, and is it getting better or worse? Include all your symptoms, even ones that seem unrelated. That odd combination of fatigue, joint pain, and a rash might point to something specific that none of those symptoms alone would suggest. And include your age, sex, and any existing conditions, since the same symptoms can mean very different things in a 25-year-old versus a 65-year-old.
Don’t stop at the top result. The correct diagnosis is much more likely to appear somewhere in the top three suggestions than as the number-one pick. Read through at least the first several possibilities and see which descriptions match your full experience. Think of the output as a shortlist to discuss with a doctor, not a final answer.
Symptoms That Shouldn’t Go Into an App First
Certain symptoms should bypass any digital tool entirely. Sudden severe chest pain, difficulty breathing, signs of stroke (facial drooping, arm weakness, slurred speech), uncontrolled bleeding, or severe allergic reactions with swelling of the throat all require immediate emergency care. Symptom checkers are designed for situations where you’re trying to figure out something that’s been bothering you, not for moments when something is clearly, acutely wrong. In emergency medicine, clinicians use “red flag” symptoms to identify conditions that need hospital-level care right away, and those situations aren’t meant to be filtered through an app first.