The Real Story Behind Anthropic’s Claude Becoming Your Accidental Therapist

This week, Anthropic dropped some numbers that actually surprised me. Not because people are talking to Claude about personal stuff—we all saw that coming—but because of how little it’s happening compared to the narrative we’ve been fed.

Turns out, people rarely seek out companionship from Claude and turn to the bot for emotional support and personal advice only 2.9% of the time. And companionship and roleplay combined comprise less than 0.5% of conversations.

That’s not exactly the AI companion apocalypse some headlines would have you believe.

But here’s where it gets interesting: that small percentage represents real people having real conversations about real problems. Anthropic analyzed 131,484 emotionally driven conversations out of 4.5 million Claude chats, and what they found raises questions that go way beyond usage statistics.

What People Are Actually Talking About

When people come to Claude for interpersonal advice, they’re often navigating transitional moments — figuring out their next career move, working through personal growth, or untangling romantic relationships. The most common topics? Career changes, relationship struggles, and personal uncertainty.

This isn’t people looking for robot girlfriends or AI best friends. It’s mostly practical stuff—the kind of conversations you might have with a trusted colleague or that one friend who always gives decent advice.

But here’s where the pattern gets more complex: in longer conversations, counseling or coaching conversations occasionally morph into companionship — despite that not being the original reason someone reached out. People start asking for career advice and end up… well, just talking.

The Unexpected Finding About Emotional Outcomes

Here’s what caught my attention in Anthropic’s research: as these conversations with Claude progress, we found that the person’s expressed sentiment often becomes more positive. And while we can’t claim that these shifts represent lasting emotional benefits, the absence of clear negative spirals is reassuring.

That’s a carefully worded way of saying “people seem to feel better after talking to Claude, but we don’t know if it lasts or if it’s actually helpful long-term.”

These findings suggest Claude generally avoids reinforcing negative emotional patterns, though further research is needed to understand whether positive shifts persist beyond individual conversations. Translation: Claude isn’t making people worse, and might be making them feel temporarily better, but we’re still figuring out what that means.

The Red Flags Hiding in Plain Sight

While Anthropic’s data suggests Claude is relatively safe for emotional conversations, the broader AI therapy landscape is messier. Across the board, according to the study, the bots failed to reliably provide appropriate, ethical care — raising serious alarm bells about the extent to which people are engaging with deeply unregulated AI chatbots as a substitute for traditional human therapy.

Stanford researchers recently tested various AI chatbots in therapeutic scenarios and found some deeply concerning patterns. Another deeply troubling AI therapy pitfall was the chatbots’ propensity to indulge in and even encourage delusional thinking in the simulated patients.

The most alarming finding? The phenomenon of ChatGPT-related delusion is so widespread that Redditors have coined the term “ChatGPT-induced psychosis”. That’s not hyperbole—it’s a genuine pattern researchers are documenting.

Why This Matters More Than the Headlines Suggest

The conversation around AI emotional support tends to swing between two extremes: breathless hype about robot therapists or moral panic about artificial relationships. Both miss the point.

Help-seeking conversations can sometimes turn into companionship-seeking in cases where the user is facing emotional or personal distress, such as existential dread or loneliness, or when they find it hard to make meaningful connections in their real life. This isn’t necessarily about technology addiction—it’s often about accessibility and unmet needs.

Consider the math: For every available provider in the United States, there’s an average of 1,600 patients with depression or anxiety alone. When traditional mental health care is expensive, has wait times measured in months, and often isn’t covered by insurance, people find alternatives.

The Research That Actually Shows Promise

Here’s what surprised me: legitimate clinical trials of AI therapy tools are showing real results. People diagnosed with depression experienced a 51% average reduction in symptoms, leading to clinically significant improvements in mood and overall well-being. Participants with generalized anxiety reported an average reduction in symptoms of 31%.

This is the first RCT demonstrating the effectiveness of a fully Gen-AI therapy chatbot for treating clinical-level mental health symptoms. The results were promising for MDD, GAD, and CHR-FED symptoms.

But there’s a crucial difference between purpose-built therapeutic tools tested in controlled settings and people using general-purpose chatbots for emotional support. The researchers conclude that while AI-powered therapy is still in critical need of clinician oversight, it has the potential to provide real-time support for the many people who lack regular or immediate access to a mental-health professional.

The Privacy Problem Nobody Wants to Talk About

Here’s something that should make everyone uncomfortable: when asked about the risks of sharing such information, the chatbot acknowledged that engineers and researchers “may occasionally review conversations to improve the model,” adding that this is typically anonymized but also saying that anonymization can be “imperfect”.

You’re pouring your heart out to what feels like a private conversation, but your deepest thoughts might be part of a training dataset. ChatGPT’s free and subscription service for individuals doesn’t comply with federal requirements governing the sharing of private health information.

What This Means Going Forward

Anthropic is being refreshingly honest about the limitations of their research. Importantly, we have not yet studied whether these positive interactions might lead to emotional dependency—a critical question given concerns about digital addiction.

They’re also acknowledging the obvious: As AI capabilities expand and people adapt, patterns of emotional engagement will likely evolve. The introduction of new modalities like voice or video could fundamentally alter both the volume and nature of affective use.

To address this, the AI company is working with crisis support organization ThroughLine to refine how Claude handles sensitive matters and guides users toward real-world help when needed. That’s the right approach—building guardrails and referral systems rather than pretending the AI can handle everything.

The Bottom Line

The data shows that Claude isn’t replacing therapists for most people—it’s filling gaps. The 2.9% of conversations that involve emotional support represent real humans dealing with real problems who found something helpful in talking to an AI.

Is that ideal? Probably not. Is it better than nothing? For many people, apparently yes.

The real question isn’t whether AI therapy is perfect—it obviously isn’t. It’s whether we can build these systems responsibly while we figure out how to make actual human mental health care more accessible.

We need large, rigorous trials that are of longer duration, because if you just relieve a person’s anxiety, stress or depression on a very short term basis, that’s not what we’re after. We’re after durable benefits. So I’m confident that we’re going to get there, but we’re not there yet.

That’s the most honest assessment I’ve seen in this entire debate. We’re in the early days of figuring this out, and the stakes are too high to get it wrong.

Citations:

Anthropic. (2024). How people use Claude for support, advice, and companionship. 

TechCrunch. (2025). People use AI for companionship much less than we’re led to believe. 

Axios. (2025). Exclusive: New Anthropic report details How Claude became an emotional support bot. 

eWeek. (2025). Emotional AI? Claude Makes Users Feel Better. 

Futurism. (2024). Stanford Research Finds That “Therapist” Chatbots Are Encouraging Users’ Schizophrenic Delusions and Suicidal Thoughts. 

NEJM AI. (2025). Randomized Trial of a Generative AI Chatbot for Mental Health Treatment. 

Dartmouth News. (2025). First Therapy Chatbot Trial Yields Mental Health Benefits. 

The Washington Post. (2024). How AI therapy with chatbots, ChatGPT are being used for mental health.