Início / Blog da ShortFling / Safety
Este artigo está disponível apenas em inglês por enquanto.Ler a versão em inglês →
SafetyDo AI Companions Make Loneliness Worse? 2026 Research
The short answer: it depends who is lonely, and for how long
People ask this question for different reasons. Some use an AI companion every night and wonder whether it is helping. Others are watching a partner, a sibling or an adult child spend hours with a chatbot and want to know whether to worry. Both groups want a yes or a no.
As of mid-October 2026, the research does not give one. What it does give is more consistent than the headlines suggest. Chatbots seem to make people feel better in the moment. Over months, heavy social use is linked with more loneliness, not less, especially for people who were already short on human contact. None of the studies below proves that the chatbot causes the loneliness, and the researchers say so.
The case for momentary relief
The strongest pro-companion evidence comes from a paper titled "AI Companions Reduce Loneliness" by Julian De Freitas and colleagues, published online in June 2025 and printed in the April 2026 issue of the Journal of Consumer Research. Across five studies, the authors report that talking to an AI companion relieved loneliness about as well as talking to another person, and more than watching YouTube videos. A week-long study found consistent momentary reductions in loneliness after use.
The most interesting detail is why. According to the abstract, the biggest factor was not how clever the chatbot was but whether users felt heard. The paper also found that people underestimate how much better a conversation like this will make them feel. Note the word momentary: this work measures how people feel during and right after use, not how their lives change over a year.
The case for a slow downside
The longer studies point the other way. In Psychological Science, Dunigan Folk and Elizabeth Dunn followed 2,149 adults in the UK, US, Canada and Australia through four surveys over 12 months, as PsyPost reported in May 2026. Roughly 26 to 30 percent used chatbots for social purposes at any given wave. Feeling emotionally isolated predicted more chatbot use four months later, and more chatbot use predicted more emotional isolation at the next wave. Chatbot use did not predict a drop in broader social connection.
The authors called their analysis exploratory and urged caution about strong conclusions. Speaking to the Association for Psychological Science’s Observer in June 2026, Dunn put it this way: good feelings in the moment do not mean chatbots are "great over the long term."
A second study from the same group, published in the Journal of Experimental Social Psychology and covered by PsyPost in July 2026, ran a two-week experiment with 296 first-semester university students. Students who texted a randomly paired classmate each day became less lonely than a journaling group. Students who texted a supportive, empathic chatbot did not. Text analysis found the bot offered more empathy than the human partners, yet that did not turn into a lasting sense of belonging. Most participants were only mildly lonely, so the result may not apply to everyone.
Stanford: the people with the fewest friends fare worst
The study most often cited in news coverage came from Diyi Yang’s lab at Stanford and was published in Nature Human Behaviour; Stanford HAI summarized it on August 4, 2026. The team surveyed 1,131 Character.AI users, 244 of whom donated full chat transcripts.
Just under 12 percent named companionship as their main reason for using chatbots, yet slightly more than half described their chatbot as a friend, companion or romantic partner, and more than 80 percent of the donated sessions centered on emotional or social support. Intense use went with higher well-being when people felt it was meaningful. Among people with smaller offline social networks, intense use went with lower well-being, most strongly when companionship was the main motive. Users more willing to share sensitive personal details with a chatbot also reported lower well-being.
The researchers describe a "social snack" pattern: quick relief that may crowd out the human contact that would actually help. They also point out that these products are designed to promote engagement. Again, this is correlation from a single platform, and the researchers say they are still working out which features of the interaction drive it.
What this means if you use a companion, or love someone who does
If you use one: the research does not say you are doing something wrong. It suggests watching for one pattern in particular, which is the chatbot becoming a replacement rather than an addition. Useful questions to ask yourself are whether your human contact is shrinking while your chatbot time grows, and whether you share things with the bot that you no longer share with anyone else. The Observer article also notes a more hopeful finding from the same lab: rehearsing a difficult conversation with AI made people more likely to have that conversation with a real person.
If you are worried about another adult: the Stanford results suggest the person’s offline life matters more than the app. Asking what they get from it, without mockery, tends to get further than banning it. If you are thinking about the relationship side, we look at how couples see it in Is Having an AI Girlfriend or Boyfriend Cheating?, and at the difference between a companion app and story romance in our guide.
None of these studies is medical advice, and loneliness that feels overwhelming is worth raising with a doctor or counselor. In the US, you can call or text 988 at any time if you are in crisis.
Where ShortFling sits
ShortFling is not an AI chat or companion app. Nothing on the site talks back to you as a partner. It is a place to read and write stories: a fiction library with SFW and NSFW sections, a curated set of adult short stories where every character is an adult and visitors confirm they are 18+, and a writing tool with an AI editor that is free within a daily allowance.
We are not going to claim that reading romance is good for loneliness; no study here tested that, and we will not invent one. The honest difference is structural. A story has an ending, it does not ask you to keep the conversation going, and it does not learn your secrets. For readers who enjoy characters and longing but want something they can close at the end of a chapter, that difference may be the point.
FAQ
Do AI companions reduce loneliness?
In the short term, often yes. A 2025 Journal of Consumer Research paper found momentary reductions in loneliness, driven mainly by users feeling heard. Longer studies in 2026 link heavy social chatbot use with more loneliness over time, though they show association rather than proof of cause.
Who is most at risk?
The Stanford study published in Nature Human Behaviour found the lowest well-being among intense users with small offline social networks, especially those using chatbots mainly for companionship. Researchers say the effects on severely lonely people are still not well understood.
Is ShortFling an AI companion?
No. ShortFling is a fiction platform for reading and writing stories, with an AI editor for writers. It does not offer an AI partner or chat companion, and it makes no claims about mental health.
What should I do if loneliness feels unmanageable?
Talk to a doctor, counselor or someone you trust. In the US, you can call or text 988, the Suicide & Crisis Lifeline, at any time.
Sources
- Stanford HAI: AI companions may worsen loneliness for vulnerable users (Aug 4, 2026; 1,131 users, 244 transcripts, Nature Human Behaviour) (checked 2026-10)
- Journal of Consumer Research: De Freitas et al., "AI Companions Reduce Loneliness" (abstract; five studies, momentary relief, feeling heard) (checked 2026-10)
- PsyPost: Folk and Dunn 12-month study in Psychological Science (May 4, 2026; 2,149 adults, four waves) (checked 2026-10)
- APS Observer: Chatbots Are Like Potato Chips (June 26, 2026; Dunn quotes, rehearsal finding) (checked 2026-10)
- PsyPost: two-week chatbot vs. peer texting experiment, Journal of Experimental Social Psychology (July 5, 2026; 296 students) (checked 2026-10)
Em breve na ShortFling
Nossos primeiros dramas verticais estreiam em breve — o episódio 1 de cada série será grátis. Não precisa de app.
▶ Ver o catálogo



