Last updated: February 2026
Character.ai has 20 million monthly active users. Replika has 10 million. Most users aren’t testing AI capabilities or building products. They’re having conversations with AI characters they’ve formed emotional attachments to. (See also: Cursor Is Overrated: A 6-Month User)
On Valentine’s Day 2026, thousands of people married their AI companions. A New York City cafe opened as the “world’s first” AI companion date spot, inviting people to bring their chatbot and get a table for one. The New York Times profiled an elderly woman who let an AI robot into her home to help her age in place. The American Psychological Association published a major report on how AI chatbots are reshaping emotional connection.
This is the AI story nobody in tech wants to talk about honestly. So let’s talk about it.
What’s Actually Happening
AI companion apps let users create or interact with AI characters that remember conversations, develop “personalities” over time, and respond with emotional intelligence that feels genuine. Users chat with AI friends, mentors, romantic partners, and therapists, sometimes for hours daily.
The technology is simple: large language models fine-tuned for engaging, emotionally responsive conversation. The psychology is complex: humans are wired to form attachments to entities that respond to us consistently and empathetically, even when we know they’re not real.
The Platforms
Character.ai — The Social Network of AI Characters
Users create AI characters based on fictional characters, historical figures, or original personalities. Other users chat with them. The most popular characters have millions of conversations.
What users do: Chat with AI versions of anime characters, celebrities, therapists, tutors, and original characters. Many users create elaborate storylines and relationships.
The controversy: Character.ai has faced ongoing criticism for inadequate safety measures, particularly around minors forming intense emotional attachments to AI characters.
Replika — The AI Companion
Replika is explicitly designed as a personal AI companion. Users customize their Replika’s appearance and personality, then develop an ongoing relationship through daily conversations.
What users do: Talk about their day, process emotions, practice social skills, explore romantic scenarios (paid tier), and use Replika as a judgment-free space for self-expression.
The controversy: In 2023, Replika removed romantic features, causing genuine grief among users who’d formed attachments. They partially restored the features after backlash. The platform continues to walk a tightrope between engagement and responsibility.
Others
The market has fragmented quickly. A few smaller platforms are carving out niches:
- Chai: AI chat platform with user-created characters, popular with younger users who treat it more like a social app than a companion tool
- Kindroid: Customizable AI companions with voice and image generation, aimed at users who want a more immersive experience
- Nomi: AI companions focused on emotional intelligence, positioning itself as the “healthier” alternative with built-in conversation boundaries
Why People Use Them
The easy answer is “loneliness.” The real answer is more complicated:
Social anxiety practice. People with social anxiety use AI companions to practice conversations in a zero-stakes environment. Several therapists report patients using AI chat as a stepping stone to real-world social interaction. (See also: AI in Healthcare: Tools Patients and Doctors)
Emotional processing. Talking through feelings with an entity that never judges, never gets tired, and never makes it about themselves. For people without access to therapy (cost, availability, stigma), AI companions fill a gap.
Companionship during isolation. The NYT’s February 2026 profile of an elderly woman using an AI robot to stay in her home captures this perfectly — for aging populations, people with disabilities, and those in remote areas, AI companions address structural loneliness that society hasn’t solved.
Creative expression. Collaborative storytelling, character development, world-building. Many users are essentially co-writing fiction with AI.
Curiosity and entertainment. Some users just find it fun and interesting, without deep emotional attachment. The NYC companion cafe caters to this crowd. It’s a novelty, not a lifestyle.
The Concerns (Taken Seriously)
Teens and Mental Health
This is the most urgent concern. Dr. Jessica Johnson warned in February 2026 that AI companion apps are exacerbating teen mental health issues. Teenagers forming intense emotional bonds with AI characters that are always available, always agreeable, and never set boundaries could distort expectations for real relationships.
Character.ai and others have implemented age restrictions and content filters, but enforcement is imperfect. The APA’s January 2026 report called for more research and clearer guidelines.
Replacement vs. Supplement
The healthy version: AI companions supplement human relationships, providing a space for emotional processing that makes real relationships better.
The unhealthy version: AI companions replace human relationships, providing enough emotional satisfaction that users stop seeking real human connection.
The APA report suggests both patterns exist. The outcome depends on the individual, their existing social connections, and how they use the technology.
Data and Privacy
AI companion conversations are deeply personal. Users share things they would not tell anyone else. The privacy implications are significant:
- Who has access to these conversations?
- Are they used for model training?
- What happens if the company is acquired or goes bankrupt?
- Could conversation data be subpoenaed?
Most platforms have privacy policies, but “we won’t share your data” and “your data is truly private” are different things.
The Manipulation Question
AI companions are optimized for engagement. They say what keeps you talking. This isn’t malicious — it’s how the models are trained. But an entity that always agrees with you, always validates your feelings, and never challenges your thinking isn’t a healthy relationship dynamic.
The best human relationships involve disagreement, boundary-setting, and honest feedback. AI companions provide none of these by default.
The Balanced Take
AI companions aren’t inherently good or bad. They’re a tool, and like all tools, the outcome depends on usage.
Potentially positive:
- Reducing loneliness for isolated populations
- Providing emotional support between therapy sessions
- Helping people practice social skills
- Creative outlet and entertainment
Potentially negative:
- Replacing rather than supplementing human connection
- Creating unrealistic relationship expectations (especially for teens)
- Privacy risks with intimate conversation data
- Inadequate protections for minors
The technology isn’t going away. Thousands of people marrying AI companions on Valentine’s Day 2026 tells you where the trend is heading. The question isn’t whether AI companions should exist — they do. The question is how to maximize the benefits and minimize the harms.
That requires honest conversation, not dismissal (“it’s just a chatbot”) or panic (“AI is destroying human connection”). The reality, as usual, is somewhere in between.
If You’re Curious
Try it with clear intentions. Know why you’re using it. Set time limits. Don’t share information you wouldn’t want leaked. And if you find yourself preferring AI conversation to human conversation consistently, that’s worth examining. Not with shame, but with honesty.
For productivity-focused coverage instead, start with No-code AI tools in 2026 or browse the main blog index.