AI may validate your feelings without considering your best interests.

Feeling understood by AI doesn't mean you're receiving sound advice.

When it comes to communication skills, using AI effectively might require questioning and refining its output.

I ran my own informal experiment using AI as my "therapist." I get it if you’re thinking, “Oh, she's a therapist—of course she's going to tell me AI therapy is bad.” Honestly, I was open to the experience, and I was surprised by the excellence of a few of the responses.

Here’s some background: I first used the AI chatbot as my emotional trashcan. Without providing identifying information, I vented my anger enough to settle my system so I could move to the next step: giving the AI a prompt addressing a real and personal communication issue I was working through.

Communication and boundary-setting are common topics in therapy. I shared some of the history of the interpersonal dynamics between this person and me, along with my current boundary-setting and communication goals, which included not setting off an emotional firestorm or inviting more back-and-forth. I then asked it what I could say.

The chatbot first provided emotional validation, which seemed excessive to me: “This sounds so painful!” followed by paragraphs of sycophancy, which is not surprising given the research on how AI tends to communicate. It took it a while to get to my actual task.

Importantly, the advice regarding what to say could have blown up the relationship I was seeking advice about. Granted, AI was textbook fabulous at validating my feelings. However, it aligned way too much with my emotions and not with my actual aims, a concern raised in some research on AI and relationships. For instance, it encouraged me to 1) say you-statements that could have easily been read as accusatory or provocative, and 2) provide too many details that would likely lead the person........

© Psychology Today