We've all been there before: headache, dizziness, sore throat. Uh-oh! Better Google the symptoms or maybe try WebMD's online symptom checker to see what's wrong.

But how accurate are these online symptoms checkers, anyway?

Turns out, millions of people are entrusting their health to some pretty lousy diagnostic systems.

Researchers tested 23 online symptom checkers and found that the correct diagnosis was provided first on a list of potential illnesses only about a third of the time. That means symptom checkers are spitting out wrong diagnoses two-thirds of the time.

"People who use these tools should be aware of their inaccuracy and not see them as gospel," says Dr. Ateev Mehrotra, who led the research and is a professor of health care policy at Harvard Medical School. "They shouldn't think that whatever the symptom checker says is what they have."

The study, published Wednesday in The BMJ, examined some of the most popular online symptoms checkers, including Ask MD, iTriage, one from the U.K.'s National Health Service and another from the Mayo Clinic.

Each website was fed 45 standard patient vignettes with symptoms consistent with medical conditions ranging from acute liver failure and meningitis to mononucleosis and a simple bee sting.

The final information provided by the symptom checkers varied. Some provided a diagnosis, while others offered recommendations for care, such as stay at home, go to a doctor or head to the ER right away.

These treatment recommendations were right on compared to diagnosis accuracy, the research team says — at least when it came to emergency situations. In cases where the patients were seriously ill, the systems accurately urged them to head to the ER 80 percent of the time.

Timely treatment, Mehrotra says, is more important than getting the diagnosis exactly right. "If a patient isn't feeling well, they're thinking, 'Is this something I have to take care of right away or can I stay home?'" he says. "We do not think distinguishing the exact problem is as important as getting to the hospital right away."

On the other hand, the researchers found that the symptom checkers were overly cautious when it came to patients who had little cause for concern, suggesting they head to a clinic when staying home and having chicken soup was appropriate.

"Using computers to help diagnose and manage care is a new frontier," Mehrotra says. "This is just the first generation [of symptoms checkers], and I'm hopeful that this research can help them improve."

iMedicalApps founder Dr. Iltifat Husain agrees and is hopeful that the mobile health world will "mature and grow up" with time. Because current symptom checkers are so new, Husain, says he wasn't surprised by the results.

"This is pretty consistent with what you see with medical-related apps out there now," he says. "More and more studies are finding that apps are not using evidence-based techniques and that they don't adequately reference medical content."

However, that doesn't stop Husain — an ER doctor at Wake Forest Baptist Medical Center — from recommending online systems and apps like symptom checkers.

"If used appropriately, these apps can help raise red flags so a patient is more proactive and seeks care in a timely fashion," he says. "Nailing the specific diagnosis right away isn't necessarily as important as nailing down, 'Do I need a health care provider?'"

Besides, Husain adds, diagnosis is what doctors are made for.

"Symptoms checkers shouldn't be for getting that final diagnosis," he says. "We spend upwards of seven years in medical school and residency to figure out how to diagnosis effectively and an application isn't going to replace that."

Copyright 2021 NPR. To see more, visit https://www.npr.org.