Please Stop Asking Chatbots for Love Advice

As he sat down across from me, my patient had a rueful expression on his face.

“I had a date,” he announced. “It didn’t go well.”

That wasn’t unusual for this patient. For years, he’d shared tales of romantic hopes dashed. But before I could ask him what went wrong, he continued, “So I asked a chatbot what I should do.”

Um. What? Simulations of human conversation powered by artificial intelligence—chatbots—have been much in the news, but I’d never had a patient tell me they’d actually used one for advice before.

“What did it tell you?” I asked, curious.

“To tell her that I care about her values.”

“Oh. Did it work?”

“Two guesses,” he sighed and turned up his hands. Although this patient was the first, it’s now become a regular occurrence in my therapy practice to hear from new patients that they have consulted chatbots before consulting me. Most often, it’s for love and relationship advice, but it might also be to connect or set boundaries with their children or to straighten out a friendship that has gone awry. The results have been decidedly mixed.

One new patient asked the chatbot how to handle the anniversary of a loved one’s death. Put aside time in your day to remember what was special about the person, advised the bot. I couldn’t have said it better myself.

“What it wrote made me cry,” the patient said. “I realized that I have been avoiding my grief. So, I made this appointment.”

Another patient started relying on AI when her friends began to wear thin. “I can’t burn out my chatbot,” she told me.

As a therapist, I’m both alarmed and intrigued by AI’s potential to enter the therapy business. There’s no doubt that AI is the future. Already, it has shown itself to be useful in everything from writing cover letters and speeches to planning trips and weddings. So why not let it help with our relationships as well? A new venture called Replika, the “AI companion who cares,” has taken it a step further and has even created romantic avatars for people to fall in love with. Other sites, like Character.ai, allow you to chat and hang out with your favorite fictional characters, or build a bot to talk to on your own.

But we live in an age of misinformation. We’ve already seen disturbing examples of how algorithms spread lies and conspiracy theories among unwitting or ill-intentioned humans. What will happen when we let them into our emotional lives?

“Even though AI may articulate things like a human, you have to ask yourself what its goal is,” says Naama Hoffman, an assistant professor in the Department of Psychiatry at the Icahn School of Medicine, Mount Sinai Hospital, in New York City. “The goal in relationships or in therapy is to improve quality of life, whereas the goal of AI is to find what is cited most. It’s not supposed to help, necessarily.”

As a therapist, I know that my work can benefit from outside support. I have been running trauma groups for two decades, and I have seen how the scaffolding of a psychoeducational framework, especially an evidence-based one like Seeking Safety, facilitates deeper emotional work. After all, the original chatbot, Eliza, was designed to be a “virtual therapist” because it asked endlessly open questions—and you can still use it. Chatbots may help people find inspiration or even break down defenses and allow people to enter therapy. But where is the point at which people become overly dependent on machines?