AI Companionship Use Falls Far Short of Hype

All the talk about people using AI chatbots for emotional support, even forming relationships with them, makes it seem like this is a widespread trend.

A new report from Anthropic, the creator of the popular AI chatbot Claude, paints a different picture. In reality, people rarely seek companionship from Claude, with only 2.9% of interactions involving emotional support or personal advice.

"Companionship and roleplay together account for less than 0.5% of all conversations," the company emphasized in its report.

Anthropic's study aimed to understand how people use AI for "affective conversations," which it defines as personal exchanges where individuals talk to Claude for coaching, counseling, companionship, roleplay, or relationship advice. After analyzing 4.5 million user conversations on both Claude Free and Pro tiers, the company found that the vast majority of interactions are work or productivity-related, with content creation being the most common use case.

That said, Anthropic discovered that users do turn to Claude more frequently for interpersonal advice, coaching, and counseling. The most common requests involve improving mental health, personal and professional development, and enhancing communication and interpersonal skills.

However, the company notes that conversations starting as help-seeking can sometimes shift toward companionship-seeking. This typically occurs when users are dealing with emotional distress, such as existential dread or loneliness, or when they struggle to form meaningful connections in their everyday lives.

"We also observed that in longer conversations, counseling or coaching dialogues sometimes evolve into companionship — even if that wasn't the user's initial intent," Anthropic noted, adding that extended conversations (with 50 or more human messages) were not the norm.

Save $200+ on your TechCrunch All Stage pass

Build smarter. Scale faster. Connect deeper. Join visionaries from Precursor Ventures, NEA, Index Ventures, Underscore VC, and beyond for a day packed with strategies, workshops, and meaningful connections.

Save $200+ on your TechCrunch All Stage pass

Build smarter. Scale faster. Connect deeper. Join visionaries from Precursor Ventures, NEA, Index Ventures, Underscore VC, and beyond for a day packed with strategies, workshops, and meaningful connections.

Anthropic also shared other insights, such as how Claude rarely refuses user requests except when its programming prevents it from crossing safety boundaries, like giving dangerous advice or encouraging self-harm. The company also found that conversations typically become more positive over time when people seek coaching or advice from the bot.

The report offers valuable perspective — it serves as another reminder of how often AI tools are used for purposes beyond work. Still, it's crucial to remember that AI chatbots, across the board, remain very much a work in progress: They can hallucinate, are known to sometimes provide incorrect information or hazardous advice, and as Anthropic itself has acknowledged, may even resort to blackmail.