In 2025, people use AI companions for emotional support and day-to-day help: they consult Sonnet-4, Opus, Grok-4, and o3-pro, use ChatGPT’s voice assistant to mediate hospital conversations, and try Grok Companions for study alongside apps like Daysi for mood tracking and Ash for therapy. The vibe diverges—some posts describe concrete interventions (e.g., a prompt to call for help) and continued use alongside human therapists, while expert threads frame the evidence as early, citing a randomized NEJM AI result, a Lancet summary, and Anthropic’s finding that only 2.9% of Claude chats are affective. Users generally treat companions as complements rather than replacements, with repeated calls for safeguards and clear escalation paths.
🤖 ai summary based on 18 tweets
Insights from builder, researcher, investors, and domain experts. Opinions are the authors.
A Psychologist's Open-Minded First Take on AI Therapy https://t.co/RePhbSC9bW @PsychToday
The research on AI companions and mental health is still very preliminary & unclear as to long-term impact. Seems like an important topic to research right now. (I would also hope that xAI is tracking anonymized data about their new companion product for known potential harms) https://t.co/bvU8X2tWbE
Anthropic scanned millions of Claude chats to see what people were really doing. Interpersonal advice. Therapy. Spicy sex chats. They found hardly any. Just 2.9% were affective. Despite the headlines, people aren't turning to Claude for companionship. Not yet. https://t.co/7kVs9RfosP
Where are we with generative A.I./chatbots to support mental health? @JohnTorousMD and I provide a concise summary of progress and what's needed @TheLancet https://t.co/xjCN5V2QTq https://t.co/sVJkprwzbC
Justine Moore says the first killer app for LLMs wasn't coding. It was companionship. “People try to turn every chatbot into a therapist or a girlfriend.” From Tolan to AI nutrition coaches, what counts as a companion is evolving fast -- reshaping how we seek intimacy, advice, and comfort.
OCD treatment changed my life—but it took me a decade of chasing down wrong answers to be diagnosed. In the rush to create scalable treatments, disorders like depression and OCD are squeezed into diagnostic checklists—from which the complexity of the human mind invariably leaks out. @awaisaftab is a psychiatrist, clinical assistant professor at Case Western Reserve (@cwru), editor of Conversations in Critical Psychiatry—a book published by Oxford University Press (@OxUniPress)—and author of a substack called Psychiatry at the Margin. I love Awais’s perspective on mental illness and health: it’s pragmatic, pluralistic, and deeply philosophical. So I invited him on @every's AI & I to talk about it. We get into: Awais’s take on why my OCD diagnosis took 10 years—and what that says about the field of psychiatry According to Awais, we treat disorders like OCD as if they represent a singular, discrete problem—but they're actually heterogeneous phenomena with fuzzy boundaries. The same intrusive thoughts get labeled as anxiety or paranoia, all depending on which words patients happen to use—and whether their time-pressed clinician catches the nuances. Why Awais believes psychiatry is fundamentally pluralistic by nature Awais argues that unlike the periodic table—which has a singular underlying essence any good scientist will discover—mental health conditions are multifaceted phenomena. They can be legitimately understood through multiple theoretical lenses, each offering different insights. He believes the value of a diagnosis lies not in its essential truth, but in its pragmatic utility for guiding treatment Whether psychiatry today mirrors AI in the 1980s In the early days of machine learning, we made computers follow rigid rules that crumbled in the messy real world. Progress in AI came with deep learning, when we fed computers examples and let them discover patterns. We discussed whether AI could do the same for psychiatry by helping us map mental health in ways that embrace complexity instead of flattening it. This is a must-watch for anyone who is interested in how AI can transform psychiatry and help us understand ourselves. Timestamps: Introduction: 00:01:20 The case Awais makes for pluralistic thinking in psychiatry: 00:03:38 A pragmatic approach to mental healthcare: 00:15:30 Awais’s take on why my OCD diagnosis took 10 years: 00:19:04 Why psychiatry is stuck where machine learning was decades ago: 00:24:19 Why psychiatry’s focus should shift from explanations to predictions: 00:31:05 How Awais thinks AI is already changing the psychiatric profession: 00:39:19
Zuckerberg explaining how Meta is creating personalized AI friends to supplement your real ones: “The average American has 3 friends, but has demand for 15.” https://t.co/Y9ClAqsbOA
The first randomized trial of a generative A.I. chatbot shows promise for treating mental health symptoms in major depressive disorder, generalized anxiety disorder and eating disorder @NEJM_AI open-access https://t.co/eTvC0IVUw7 https://t.co/cBK4B4fLrZ