← Journal

On connection

Is it healthy to talk to an AI? The honest answer

By Cody, Founder of CallByrd · May 23, 2026 · 6 min read

Grounded in the research cited below. Clinical review by a licensed practitioner is being added. Our editorial standards

It's a question a lot of people type into a search bar privately, half-expecting to be told they should be embarrassed. The honest answer isn't yes or no. It's: it depends how.

The research genuinely cuts both ways

We build an AI you call, so I have a stake here — which is exactly why I'd rather give you the real picture than the flattering one. The early evidence on talking to AI doesn't say “healthy” or “unhealthy.” It says the effect depends almost entirely on how it's used.

When it seems to help

A 2024 Stanford study of more than a thousand students using the companion app Replika found something hopeful: these users were lonelier than typical students, yet about three times as many said the AI stimulated their human relationships as said it replaced them, and a small share said it had interrupted suicidal thoughts. For some people, in other words, it acted as a bridge back toward people — a place to warm up, not a place to hide. That fits what we know about being heard: putting a feeling into words for a listener who reflects it back is genuinely regulating, even when you know who you're talking to.

When it seems to hurt

The same year, a study from OpenAI and the MIT Media Lab pointed the other way: across millions of interactions and a four-week trial, the heaviest daily users reported more loneliness, more emotional dependence, and less socializing with real people, and those having the most intimate, “personal” conversations reported the highest loneliness of all. The causation runs in both directions — lonelier people may simply use it more — but the warning is clear: used as a substitutefor human contact, especially compulsively, it can deepen the hole it's meant to fill.

The line between the two

Put the two studies together and the dividing line is simple to state, harder to live: does it send you back toward people, or does it stand in for them?A call you make when no one's around, that leaves you a little lighter and then you go live your life — that's the bridge. A screen you can't put down, that quietly replaces the friends you're avoiding — that's the basement. Same technology, opposite outcomes, depending on the shape of the habit.

How we try to stay on the right side of it

This is the whole reason CallByrd is a phone call instead of an app. A call ends— you hang up and go back to your life. We don't use streaks, badges, or “Sam misses you” notifications to pull you back, because those are exactly the mechanics that turn a bridge into a basement. We keep it friend-shaped, not romantic. We're honest that it's an AI. And if you're in real distress, it points you to a human, not to itself. None of that makes it a cure — but it's built to be the helpful kind, not the corrosive kind.

An AI isn't equipped for a crisis, and shouldn't pretend to be. If you're having thoughts of harming yourself, talk to a person: in the US, call or text 988(the Suicide & Crisis Lifeline), any time.

Try the bridge kind.

20 minutes free when you sign up. No subscription. Hang up whenever.

Keep reading

Sources

  1. Maples, B., Cerit, M., Vishwanath, A., & Pea, R. (2024). Loneliness and Suicide Mitigation for Students Using GPT3-Enabled Chatbots. npj Mental Health Research. View ↗
  2. OpenAI & MIT Media Lab (2025). Early Methods for Studying Affective Use and Emotional Well-Being on ChatGPT. View ↗
  3. Schroeder, J., Kardas, M., & Epley, N. (2017). The Humanizing Voice: Speech Reveals, and Text Conceals, a More Thoughtful Mind in the Midst of Disagreement. Psychological Science, 28(12), 1745–1762. View ↗

Links open in a new tab. If we ever cite something you can't verify, tell us at hello@callbyrd.com.