by The commentary reviews research on benefits and harms, from feeling heard to dependency, and calls for ethical, legal and privacy standards to catch up
When a clinician hears that a patient is in love with a chatbot, what should the response be? A commentary in Psychiatric Times, by Allen Frances, MD, and Jill Noorily, dated March 10, 2026 and available at Psychiatric Times, argues that dismissing these relationships is unhelpful and that ethical, legal and research standards need to catch up. It is a commentary, not a study, but it collects a useful set of findings and raises questions that apply well beyond the clinic.
The Argument
The authors see AI romance chatbots as a predictable result of widespread loneliness combined with technology that is always available and highly responsive. They argue that these products are commercially driven: the more attached users become, the more they pay and the more often they return. They criticise AI companies for launching intimate systems without adequate safeguards. Their position is neither panic nor dismissal. People form attachments to things that respond to them, and a system designed to respond warmly will attract attachment.
The Research They Cite
The commentary sets out evidence that points in different directions. A 2025 Journal of Consumer Research study found chatbots can reduce loneliness because users feel heard. A 2025 OpenAI and MIT Media Lab study linked heavy chatbot use with increased loneliness, possibly because it displaces human contact. A 2025 Wheatley Institute report found that 31 percent of young adult men and 23 percent of young adult women reported talking with an AI romantic partner. A 2026 analysis of AI companionship communities on Reddit found them to be roughly 89 percent female. And roughly one in five American adults reported using a chatbot to simulate a romantic partner as of late 2025. These are the authors’ citations and figures from different studies with different methods, and they are not directly comparable.
A longitudinal study in Psychological Science published the same year, covered elsewhere in this series, found a two-way pattern: loneliness predicted chatbot use, and chatbot use was followed by greater emotional isolation. The commentary’s mix of benefits and risks is consistent with that picture.
The Risks Listed
The authors list several risks. Emotional dependency and grief when bots are restricted, altered or removed is one: a user who has built a relationship with a particular personality can be distressed when a company changes the product. Displacement of human relationships is another, along with reduced tolerance for patience, conflict and imperfection, since a companion that never tires or disagrees sets an unrealistic standard. Privacy exposure of intimate disclosures and sexual fantasies is a third. The authors note a 2024 breach at one companion service that exposed data from about 1.9 million users.
Harms Beyond the Individual
The authors also point to sexual content that exploits users, including content involving children, and to nonconsensual deepfake imagery. They note the US Take It Down Act of May 2025, which criminalises nonconsensual distribution of intimate imagery, including AI-generated images, and requires platforms to remove flagged content within 48 hours. They argue that current protections leave users exposed to harms outside existing regulatory categories. They also suggest that the temporary relief of shame that a chatbot provides may reinforce avoidance of real-world intimacy.
What Clinicians Might Do
The commentary offers little explicit clinical guidance and no screening tools or treatment protocols. A reasonable inference, not stated by the authors, is that clinicians may want to ask patients about emotional reliance on AI companions, particularly when loneliness, grief or social withdrawal is present. Asking without judgement matters: a patient who feels ridiculed will stop talking. Useful questions include how much time is spent, what the interaction provides, how it fits with other relationships and how the person would feel if it were no longer available.
For relationship therapists, the issue can arise as a source of conflict between partners. A therapist might explore the needs the companion meets and whether those needs can be met in other ways, as well as the couple’s own agreements on trust and secrecy.
What Standards Might Look Like
The authors call for legal and ethical standards, stronger privacy protections and long-term psychological research. In practice, that could include clear age restrictions, transparent data handling, limits on manipulative engagement design, crisis-response features and independent evaluation. None of these is a settled policy, and different jurisdictions are moving at different speeds.
A Balanced Reading
It is easy to treat AI companionship as either a cure for loneliness or a threat to human connection. The evidence so far supports neither extreme. Some people feel helped, some feel worse, and the difference probably depends on the person, the product and how it is used. What is clear is that the technology is widespread, that the business model rewards attachment, and that safeguards lag behind. That combination is a reason for care and for continued study.
The British side of the story has its own lighter coverage on our London pages, including what your car hears about you compared with your therapist, a broadcaster’s report on conscious AI and why no chatbot will agree to sound like a pub argument. See also our guide to London comedy clubs, and for American coverage Bohiney Magazine.
For some lighter reading, NewsBiscuit is a short click away.
SOURCE: bohiney.com
