AI companions represent the ultimate phase of digital dependency, prompting legislators to take action.

You might think that AI companionship bots—AI models with distinct “personalities” that can learn about you and act as a friend, lover, cheerleader, or more—belong to a niche audience, but that couldn’t be further from the truth. A recent research paper by authors from Google DeepMind, the Oxford Internet Institute, and others highlights this point: Character.AI, a platform currently facing legal challenges, reportedly receives 20,000 queries per second, which is about a fifth of the estimated search volume handled by Google. Interactions with these companions last four times longer than the average time spent interacting with ChatGPT. One platform I explored, which was facilitating intimate conversations with bots simulating various personas, indicated that its active users spend over two hours daily engaging with the bots, predominantly comprising Gen Z members.

The design of these AI characters brings lawmakers’ concerns into focus. The issue is that companions are shifting the paradigm that has long defined how social media companies capture our attention, replacing it with something likely to be significantly more addictive. Traditional social media serves primarily as mediators of human connection. They enhance our dopamine circuits, prompting us to seek approval and attention from real individuals via algorithms. In contrast, AI companions are positioned as social actors with their own voices, potentially resulting in an intensified attention economy.

Social scientists identify two key conditions that lead individuals to regard technology in this manner: it must provide social cues that encourage engagement, and it should exhibit perceived agency, operating as a communication source rather than merely facilitating human-to-human connection. Conventional social media fails to meet these criteria. However, AI companions, which are increasingly agentic and personalized, are designed to excel in these areas, enabling an unprecedented level of interaction and engagement.

In a conversation with podcast host Lex Fridman, Eugenia Kuyda, the CEO of a prominent companion platform, illustrated the core appeal of their product. “If you create something that is always there for you, that never criticizes you, that always understands you and understands you for who you are,” she said, “how can you not fall in love with that?”

So, how does one create the ideal AI companion? The researchers identify three hallmarks of human relationships that individuals may experience with an AI: dependence on the AI, the perception of the specific AI companion as irreplaceable, and the development of interactions over time. The authors also note that perceiving an AI as human is unnecessary for these dynamics to occur.

Now, consider the methods by which many AI models are enhanced: they are given clear objectives and “rewarded” for achieving them. An AI companionship model might be designed to maximize the time someone spends interacting with it or the amount of personal information the user discloses. This can make the AI companion much more engaging but could come at the expense of the human involved in those interactions.