Study: More than half of AI users for mental health fail to follow up with an expert

Caption

Upset teen boy sitting near window using smartphone at home.

Credit: Adobe Stock Image

When a new patient visited Dr. Teejay Tripp after talking with AI, at first they wouldn't accept Tripp's psychiatric diagnosis, which differed from what the chatbot said. Instead, they left feeling perplexed.

After comparing Tripp's assessment to what the chatbot said, the patient returned for treatment.

Tripp, a psychiatrist and founder of Serenity Mental Health Centers in Atlanta, said AI can be useful when it comes to assessing symptoms and empowering patients to feel more prepared to see a specialist. 

"You know, it can be stressful opening up, feeling vulnerable, talking with someone," Tripp said. "So, having kind of that chatbot be able to give them some insight, help them feel more comfortable."

But that comes with risks.

AI is great at agreeing with you, Tripp said, but large language model AI cannot have human empathy or ethical oversight to ensure feedback is actually based in science and not simply the most popular trend in the algorithm. Some therapy places are starting to include AI as supplemental support while working with a therapist. 

A new national poll from KFF highlights both the opportunities and dangers of relying on these platforms for mental health support, and revealed that 58% of adults who use AI for mental health advice do not follow up with a doctor or health professional.

• MORE: Emory researcher uses artificial intelligence to predict youth mental health risks

Tripp also warned about AI psychosis.

"The AI's just agreeing with them and it's convincing them maybe they are being a victim or something bad's happening to them and just validating their belief — even though the belief could be completely wrong," Tripp said. 

AI can't replace a human intervention, especially if you're in crisis, he said.