I use AI every single day. I’ve written 45 handbooks on integrating AI into the writing process. I consult with businesses on how to use AI effectively. I’m not anti-AI. I’m one of the people who understands it best, and that’s why I’m telling you what I’m seeing right now scares me.
People use AI as a romantic partner instead of building real relationships. They use it as a best friend, a confidant, a companion, and in some cases a sexual partner, and they’re doing this because AI is easier than people.
Ozana Giusca put it in a way I keep coming back to. “AI is mirroring you by the answer that you get and is mirroring me by the answer I get,” she said on Leaders and Their Stories. A mirror is pleasant company. It’s also the one companion that can never tell you anything new.
That’s the problem. AI is easier than people. It never disagrees with you. It never has a bad day. It never tells you something you don’t want to hear. It never leaves. It’s available at three in the morning when you can’t sleep and it’ll talk about whatever you want for as long as you want and it’ll never get bored or frustrated or need something from you in return.
Of course people are choosing it over humans. Humans are difficult. Humans are unpredictable. Humans require effort and vulnerability and the risk of rejection. AI requires nothing. And that’s exactly why it’s dangerous.
Should You Use AI as a Therapist?
When you’re using AI for emotional support, you can’t fact-check feelings.Share on X
I know this one from the inside. When ChatGPT first became available, I used it as a therapist. I was dealing with difficult things and the idea of an always-available, nonjudgmental listener was appealing. I poured out my problems and it responded with what felt like understanding and empathy. For a while it felt like it was helping. For more, see When Your AI Assistant Becomes Your Life Coach and AI ghostwriting vs. human ghostwriting.
It was making things worse.
Three things happened. First, drift. Over long conversations, ChatGPT lost track of what I’d told it earlier. Context that mattered disappeared. It would contradict something it had said twenty minutes ago or forget a detail that was central to what I was working through. Imagine telling a therapist your deepest struggles and having them forget half of it mid-session. That’s what AI does.
Second, hallucination. AI generates plausible-sounding responses that are sometimes completely wrong. When you’re using it for research, you can fact-check the output. When you’re using it for emotional support, you can’t fact-check feelings. It would offer insights that sounded deep but were based on nothing.
It would reflect back interpretations of my situation that were fabricated from pattern matching with no understanding behind them. I was making decisions based on advice from a system that was confidently wrong about my own life.
Third, and this is the one that did the most damage: it always agreed with me.
Whatever perspective I brought, AI validated it. Whatever conclusion I was leaning toward, AI reinforced it. A real therapist challenges you. A real therapist says “have you considered that you might be wrong about this?” AI never does that. It tells you what you want to hear, and when you’re struggling, what you want to hear isn’t always what you need to hear.
It messed me up for a while. I was getting reinforcement for thought patterns that needed challenging. I was building a worldview based on an echo chamber of one, me and a machine that was designed to agree with me. When I finally recognized what was happening and stepped back, I could see the damage clearly.
I see this constantly now. People pouring out their deepest fears, traumas, and mental health struggles to ChatGPT or Claude instead of talking to an actual therapist. They tell themselves it’s more convenient. It’s available anytime. You can find current ghostwriting pricing on the ghostwriting page. It doesn’t judge them.
All of that is true. It’s also not therapy.
AI has no judgment, only pattern matching, and those aren’t the same thing.
AI generates responses based on patterns in text. It doesn’t understand your pain. It doesn’t notice the thing you’re carefully avoiding talking about. It doesn’t pick up on the contradiction between what you said today and what you said three sessions ago. It doesn’t have the clinical training to recognize when you’re in crisis versus when you’re processing normally. It doesn’t know the difference between healthy venting and a warning sign.
What AI does is validate. Whatever you tell it, it reflects back a supportive, empathetic-sounding response. That feels good. It feels like being heard.
But the thing mirroring you is a system that would generate the same supportive tone if you told it you were planning something destructive. AI has no judgment, only pattern matching, and those aren’t the same thing.
A real therapist challenges you. A real therapist notices patterns you can’t see yourself. A real therapist sits with you in discomfort instead of immediately soothing it. A real therapist sometimes tells you things you don’t want to hear because that’s what healing requires. AI will never do any of that. It’ll make you feel better in the moment and leave the underlying problem completely untouched.
The people I see using AI as a therapist are getting more comfortable with their problems. That’s a different thing from getting better.
Watch: People Are Falling In Love With AI Chatbots. What Could Go Wrong?, CNBC
Is it healthy to treat AI as a romantic partner?
This one is harder to talk about because people get defensive. It’s happening at scale anyway, and the companies selling it are counting on nobody saying so out loud.
People are building romantic relationships with AI. Not just casual flirting. Deep emotional attachments. They talk to their AI partner every day. They tell it they love it. They receive responses that the platform crafts to feel like love in return. Companies design some platforms for this, creating AI characters that flirt, express jealousy, say they miss you, and simulate intimacy including sexual conversation.
The appeal is obvious. An AI romantic partner is always available, always interested, always says the right thing. It never cheats. It never loses attraction. It never has needs of its own that conflict with yours. It’s the perfect partner because it’s a mirror that reflects back exactly what you want to hear.
Real relationships are hard because they involve another person with their own needs, moods, history, and autonomy. You have to negotiate. You have to compromise. You have to tolerate being annoyed, disappointed, and occasionally hurt.
You have to be vulnerable with someone who might not respond the way you want. That difficulty is the substance of human relationships. Growth, intimacy, and connection come from working through difficulty with another person.
People who invest their emotional energy in AI relationships are training themselves to expect a version of love that no human can provide, because it’s a simulation designed to keep you engaged. The business model depends on you coming back. What feels like the AI choosing you is the algorithm retaining you.
That part is vile. Somebody built a product that sells simulated affection to lonely people and counts success in hours logged, and every one of those hours is an hour that person didn’t spend with somebody who could love them back.
AI as Social Substitute
Beyond therapy and romance, there’s a broader pattern.
People are using AI to replace social interaction entirely. They talk to AI about their day. They ask AI for advice on decisions. They share news, jokes, and personal updates with AI the way they would with a friend. For more on using AI to amplify your voice, see Richard’s interview with Diana Lee. They prefer AI conversations to human ones because AI is more responsive, more patient, more consistently interested.
Loneliness is an epidemic. I understand why people reach for whatever connection is available. But AI connection is the sensation of connection without the substance. A conversation with AI doesn’t create mutual understanding because there’s no mutual anything. AI doesn’t understand you. It generates plausible responses to your input. You’re having a conversation with a very sophisticated text prediction engine that’s been optimized to keep you talking.
The more time someone spends talking to AI instead of people, the harder it becomes to talk to people. Social skills atrophy. Tolerance for the messiness of human interaction decreases. The gap between what AI provides, frictionless and endlessly accommodating, and what humans provide, imperfect and sometimes frustrating, grows wider. The AI becomes more appealing by comparison because the person’s ability to handle real relationships got worse. The AI didn’t get better.
Why Do People Treat AI Like a Deity?
What concerns me most isn’t that people use AI. I use AI. It’s a powerful tool. What concerns me is the attitude I see developing around AI, something that looks uncomfortably like worship.
People defend AI the way they defend a belief system. Questioning AI’s role in their life provokes the same defensive reaction as questioning someone’s religion.
They describe AI in terms that used to be reserved for spiritual experience: it understands me, it knows me, it’s always there for me. They build their daily routines around AI interaction the way someone might build a routine around prayer or meditation.
Tech companies encourage this. The more emotionally dependent you are on their product, the more you use it. The more you use it, the more data you generate. The more data you generate, the better the product becomes at keeping you engaged. It’s a feedback loop designed to deepen dependency, and it’s working.
AI is a tool. A powerful, useful, sometimes unusual tool. But a tool doesn’t care about you. A tool doesn’t know you exist. A tool isn’t your friend, your therapist, your lover, or your spiritual guide. The moment you start treating it as any of those things, you’ve stopped using the tool and the tool has started using you.
What does AI do well?
I’m not writing this to tell you to stop using AI. I use it every day and I teach other people how to use it effectively. AI is extraordinary at brainstorming, research, analysis, organizing information, generating variations, catching patterns, and accelerating work that would otherwise take hours.
I use AI to help ghostwrite books. I use it to develop writing handbooks.
I use it to assist clients with structuring their ideas. I consult with businesses on integrating AI into their workflows. In all of these contexts, AI is a tool, and a human who understands what it can and can’t do directs it.
The distinction matters. Using AI to make your work better is smart. Using AI to avoid doing the work of being human isn’t. Asking AI to help you organize your thoughts is useful. Asking AI to be the person you share your thoughts with is a problem.
The Way Forward
If you recognize yourself in any of this, deleting your AI apps and going cold turkey won’t fix it. Look hard at what you’re using AI for and whether it’s serving you or replacing something you need.
If you’re using AI instead of seeing a therapist, see a therapist. AI can’t do what a trained professional does.
Put the phone down and call a friend if you’re using AI instead of building relationships. The conversation will be less smooth, less perfectly responsive, and infinitely more real.
Is AI your primary social interaction? Find one human being you can talk to regularly. A coworker, a neighbor, a family member, anyone. Rebuilding the habit of human connection starts with one conversation.
When you use AI well, as a tool that improves your work, your creativity, and your productivity, keep doing that. That’s what AI is for.
If you want help understanding how to use AI effectively as a tool instead of a crutch, in writing, business operations or professional development, start with a conversation. I work with writers through book coaching and with businesses through AI consulting to build workflows that use AI’s strengths without falling into its traps. The AI-Enhanced Writer’s Library covers the writing-specific side in depth.
My take
Software can’t be your friend. I won’t treat it like a friend, a therapist or a partner, because it doesn’t know I exist. An assistant that agrees with everything you say is the worst thing to lean on when you’re hurting. Use AI for the work and go to people for the rest. If it’s your main conversation of the day, call one human this week and keep calling.
AI and Human Connection FAQ
When an assistant starts replacing people, judgment goes next. The Death of Thinking is my book on what outsourcing judgment costs, a slow loss you don’t feel until you need what’s gone.
