AI Romance Draws 28% of U.S. Adults as Anthropic President Warns Against Chatbot Partners
AI companionship is already part of how some people seek intimacy, even as Anthropic co-founder and president Daniela Amodei argues that romantic relationships are a poor use of the technology. A survey released last year found that 28.16% of U.S. adults had at least one intimate or romantic relationship with AI. That puts the debate beyond a hypothetical product use case, though the figure doesn’t show how respondents defined a relationship or how long it lasted.
The survey, conducted by Vantage Point Counseling Services among 1,012 adults, found that 53.95% had formed some kind of relationship with AI. The broader category may include experiences that differ sharply in depth and meaning, so it shouldn’t be read as a count of people in committed partnerships with chatbots. Still, the findings suggest that interacting with AI in emotionally significant ways has reached a meaningful share of the public.
The limits of the evidence matter. The study describes respondents’ reported experiences; it doesn’t establish whether those relationships help or harm users, or how they affect social connections over time. The numbers here are striking, but they answer a narrower question than the public debate often assumes.
Amodei draws a line around AI companionship
Speaking in a conversation recorded for the Think Health conference on Friday, Amodei said people shouldn’t use AI as a romantic partner. “I think that’s a terrible use of AI and people should not do it,” she said. Her view is unusually direct in a discussion where companies often describe conversational systems in broad terms such as assistants or companions, leaving users to decide what those labels mean in practice.
Her concern is that AI romance could extend a broader shift toward spending more time with technology and less time with other people. She cited platforms such as Snapchat and Instagram as examples of digital services that can displace face-to-face conversation. AI partners, in her view, could take that substitution further by offering software as a stand-in for human companionship. The concern isn’t simply that people spend time on screens; it’s that a system designed to respond on demand may make the friction and unpredictability of human relationships easier to avoid.
That possibility deserves scrutiny, but the survey doesn’t show that AI relationships have displaced human ones. A person might use a chatbot alongside friendships and family ties, or turn to it when other support isn’t available. The available figures don’t distinguish among those situations. And this is the part most people overlook: a reported relationship with AI can describe a wide range of experiences, from occasional emotional conversation to a sustained romantic attachment.
“Those are fundamental human relationships that I actually think will become more important,” Amodei said, referring to roles such as manager, therapist and partner. She also acknowledged that skepticism about technology’s ability to bring people closer is warranted. “It would probably be a better world if we were spending less time interacting with technology and more time interacting with each other,” she said. Her argument rests on a judgment about what people need, not on evidence in the survey that any particular use of AI has caused social harm.
That position sets a clear boundary for Claude, Anthropic’s chatbot, even as the company builds and sells AI products. Anthropic told investors earlier this week that it generated $16.8 billion in third-quarter revenue. That commercial context sharpens the question around the company’s public stance: the business benefits from broader AI adoption, while its president is warning against one possible consumer use. The company’s financial results and her comments sit alongside a consumer question with no settled answer: how users will treat systems that can simulate attention and affection.
AI as a way to return time to people
Amodei’s preferred role for AI is to reduce repetitive or isolating work and give people more time for friends, family and partners. In healthcare, for example, she said AI could help doctors find information faster, leaving more of an appointment for conversation with patients. The value in that example depends on what happens to the time saved: faster information retrieval matters to the patient only if it makes room for more attentive care.
That distinction is central to her argument: AI can support human relationships by making work more efficient, but it shouldn’t take the place of those relationships. It’s a sensible test, though it isn’t automatic. A tool can save time on one task while drawing people into more screen-based interaction elsewhere. If you’re working in this space, the practical question is whether a product helps users connect with other people or encourages them to treat the system itself as the destination.
The survey data show that many adults have already explored an emotional connection with AI, while Amodei’s remarks point to a different measure of progress: whether the technology gives people more time to connect with one another. That’s a harder outcome to assess than adoption, but it speaks more directly to her stated concern. The difference between assistance and substitution may depend on how a system is used, what it promises, and whether it steers users toward human support when that’s what they need.
What the findings leave unresolved
Vantage Point’s researchers said the direction is uncertain. The share could rise as AI advances, they wrote, or decline if users encounter its limits. Neither possibility is established by the survey, and the findings don’t tell us what product changes or user experiences would drive either outcome. For companies developing chatbots, the figures point to an audience already experimenting with emotional connection, but they don’t settle what users want from these systems or where people draw the line between assistance and companionship.
That uncertainty is more significant than it looks. The central issue isn’t whether a chatbot can produce affectionate language; it’s what role users come to assign it in their lives. For now, the survey records reported behavior and Amodei offers a clear view of the boundary she believes technology should respect. Whether users share that view remains open.