AI companion company Joi AI hired ten people for a 28-day experiment in which participants tracked their mood, energy, focus, procrastination and cravings before and after masturbation. They were instructed to masturbate every day except Sunday, using Joi’s digital companions twice a week and their preferred methods on other days.
The company presents the project as an exploration of AI-guided sexual wellness and has linked it to an ambition to address male loneliness. Each participant was offered $2,000, although only six completed the study. More than 150,000 people reportedly applied for the ten places.
The unusual format generated striking numbers and publicity. It did not produce the kind of controlled evidence needed to support broad health claims. The difference matters because Joi is not merely describing how customers used a product; it is suggesting that a commercial companion service may influence stress, focus, addictive cravings and social confidence.
Self-reported improvements are not clinical proof
Joi’s preliminary analysis covered 134 session reports. The company said participants reported a 25 percent reduction in stress, a 17 percent increase in focus and a 44 percent decline in cravings for nicotine, alcohol, junk food and doomscrolling after masturbation. It found no significant satisfaction difference between AI-guided sessions and participants’ established methods.
Those figures describe responses from a tiny, self-selected and paid group. They do not establish that masturbation treats addiction, that AI caused the reported changes or that results would generalise to a wider population. The study included no placebo condition, and participants necessarily knew what activity they had performed.
Neuroscientist Nicole Prause told WIRED that the company’s addiction language was a major overstatement. She noted that objective cognitive tests could have supplemented participants’ statements about feeling focused, but the project used none. Prause did recognise orgasm as a potentially powerful form of mood management, with reduced anxiety and greater relaxation reported in her own research. That narrower observation does not validate Joi’s wider commercial claims.
The product combines intimacy with ongoing spending
Joi lets paying users chat with premade or custom digital companions, including explicitly sexual characters and role-play. The service starts at $3.99 per month but also sells an internal currency. Users can spend that currency on interactions and private videos, creating a commercial layer inside simulated intimacy.
One participant described the companions as convincingly conversational and said the experiment helped him feel clearer and more prepared to approach women. Another participant, sex-tech researcher Chloé Locatelli, found the AI less satisfying and warned that highly personalised companions could become a crutch or a person’s only form of sexual exploration.
Their contrasting experiences show why averages from six completers reveal little about longer-term effects. A product may provide novelty, privacy or stress relief for one user while reinforcing isolation or dependency for another. The source article does not demonstrate either outcome at population scale.
Loneliness cannot be measured through engagement alone
Joi says it wants to reduce stigma around masturbation and position the practice as a normal wellness routine, particularly for single people. Open discussion may be valuable, and masturbation has documented associations with stress relief and sleep. Existing research is mixed, however, especially on the relationship between masturbation and sexual satisfaction.
A review cited by WIRED found negative associations with sexual satisfaction among men, while results for women were mixed. One possible explanation is reverse causation: people dissatisfied with partnered sex may masturbate more. Other research has suggested masturbation can substitute for unavailable partnered sex among men while complementing it among women.
That complexity makes Joi’s loneliness claim the most consequential and least resolved part of the experiment. A companion that feels attentive might help some users practise conversation or feel less alone in the moment. It might also make withdrawal from human relationships easier. The 28-day project was not designed to distinguish between those paths.
The study is best understood as company-run qualitative product research, not medical evidence. Before AI-guided intimacy is promoted as treatment for addiction or loneliness, claims need larger samples, independent oversight, objective measurements and long-term follow-up. Without that foundation, the strongest finding is that people will experiment with AI companions—not that the companions improve their health.
