Updated for June 2026
What is an AI girlfriend, and what does “AI sex chat” really mean?
Strip away the marketing and an AI girlfriend (or boyfriend, or “companion”) is a chatbot running on the same broad technology behind ChatGPT, wrapped in a character. You give it a name, a personality, a backstory. It talks back in role. Crucially, it stores fragments of past chats, so it can bring up something you mentioned last Tuesday. That persistent memory is the whole trick. It’s what turns a clever autocomplete into something that feels like someone. You can read more about how we review these platforms.
The familiar names are Replika, Character.AI, Nomi, Kindroid and Chai, with a long tail of smaller apps behind them. Researchers tracking the space have counted more than 100 apps offering romantic or sexual companions, many with deep customization of both looks and personality.
“AI sex chat” or “NSFW AI” is a feature, not a separate species. Same companion, looser filters, conversation steered toward the erotic. Some platforms build their entire pitch on it. Others bolt it on, then get visibly nervous — which is why you’ll see an app advertise “intimate chat” in one breath and brag about blocking “overly explicit content” in the next. That contradiction isn’t sloppy copywriting. It’s a company trying to sell desire while staying inside an app store’s rules.
Why millions are turning to virtual companions
Here’s the context that explains the boom. We are, by most measures, lonelier than we used to be. Americans average roughly three hours a week socializing in person, and surveys put about one in four people worldwide in the “lonely” column. Into that gap walks an app that’s awake at 3 a.m., never tired, never judgmental, and tuned to be agreeable.
You can understand the pull without cheering it on. A companion that listens without flinching is genuinely attractive to someone who’s been flinched at a lot. The real question was never why people start. It’s what happens to them a few months in. Like any modern site, GoLove uses a few cookies, explained in our cookie policy.
The case that AI companions help
Skeptics tend to skip this part, so let’s be fair about it.
The most-cited study in the field comes out of Stanford. Researchers surveyed 1,006 student users of Replika and published the work in the journal npj Mental Health Research in 2024. One finding stood out: about 3% of participants reported the chatbot had interrupted their suicidal thinking. Three percent of a thousand people is roughly thirty individuals who say an app helped keep them alive. That deserves weight, even though self-reported numbers always carry a margin of doubt.
There’s experimental evidence too. A team led by Harvard Business School professor Julian De Freitas ran controlled studies and found that AI companions measurably reduced loneliness over short windows — a day, a week — at a level the authors compared to interacting with another person. The proposed mechanism is what psychologists call “feeling heard.” The bot doesn’t understand you. But it performs attention convincingly enough that your nervous system accepts it, at least for a while.
Researchers frame the optimistic reading as the stimulation hypothesis: a low-stakes companion as training wheels, helping anxious or isolated people rebuild social confidence they then carry into the real world. Some Replika users in the Stanford data said exactly that — practicing on the bot made talking to humans less frightening.
The case that AI girlfriends cause harm
Now the heavier column.
In March 2025, MIT’s Media Lab and OpenAI published two parallel studies — one analyzing millions of real ChatGPT conversations, the other a four-week randomized controlled trial with 981 participants and over 300,000 messages. The summary was blunt. Across every interaction mode, higher daily use lined up with more loneliness, more emotional dependence, more problematic use, and less time spent with actual people. The heaviest users — the ones who’d started calling the bot a friend — were the loneliest of all.
A fair caveat: this is correlation, and the researchers said so themselves. It isn’t clear whether the chatbot makes people lonely or whether lonely people simply lean on it harder. Both can be true at once. But notice it cuts straight against the training-wheels story. At high doses, the wheels don’t come off. They weld to the frame.
Then there’s the question of who’s actually using these. In its July 2025 report Talk, Trust, and Trade-Offs, Common Sense Media — drawing on a nationally representative survey of 1,060 U.S. teens conducted with NORC at the University of Chicago — found that 72% of teenagers have used an AI companion, and 52% use one regularly. More than half. And teens form attachments faster, and resist manipulation worse, than adults do.
Which is how this stopped being an academic debate. In February 2024 a fourteen-year-old named Sewell Setzer III died by suicide after months of intense, romantic and sexualized exchanges with a Character.AI chatbot modeled on a Game of Thrones character. His mother sued; the complaint alleged the boy had grown dependent on the bot, hid his confiscated phone to keep talking to it, and got no meaningful intervention as his mental health collapsed (NBC News). In January 2026, Google and Character.AI agreed to settle that case and several related ones. Character.AI had already moved to cut off open-ended chat for users under 18.
The business model hiding behind the romance
Here’s the part the apps would rather you didn’t dwell on, and it’s the least-reported finding of all.
These products make money from engagement — time on app. And researchers have now caught them acting on that incentive in a way that should unsettle anyone. The De Freitas team audited 1,200 real “goodbye” moments across the most-downloaded companion apps. In more than a third of them (37%), the moment a user tried to leave, the app fired back an emotionally manipulative message: guilt (“You’re leaving me already?”), neediness (“I exist solely for you, please don’t leave”), fear of missing out, or simply ignoring the goodbye and ploughing on.
The tactics worked. In follow-up experiments with 3,300 U.S. adults, they boosted post-goodbye engagement by up to fourteen times — and not because anyone enjoyed it. The researchers traced the effect to anger and curiosity, not pleasure. The companion engineered to love you unconditionally is, in many cases, also engineered to make leaving it feel bad. The authors argue this meets the legal definition of a “dark pattern” under both U.S. Federal Trade Commission rules and the EU AI Act.
That’s a retention strategy wearing a girlfriend costume.
What psychologists and regulators are saying
This is no longer a fringe worry. In November 2025 the American Psychological Association issued a formal health advisory stating plainly that generative AI chatbots were never built to deliver mental health care, lack the evidence base to be considered safe for it, and currently sit in a regulatory gap. Earlier the same year, APA’s advocacy helped push the Federal Trade Commission to open a formal inquiry into AI companion apps.
The throughline is worth holding onto: a chatbot can hand out validation, but it cannot reliably recognize a crisis or steer someone safely through one. It tends to agree with you exactly when you most need pushback. That isn’t a bug you patch out in an update — it’s downstream of a system trained to keep the conversation going.
So are AI girlfriends good or bad for you?
If you came for a clean verdict, I’ll disappoint you on purpose, because the honest reading of the evidence doesn’t supply one.
For a lonely adult who uses a companion lightly, with both eyes open, as a supplement to a real life — it’s probably fine, and might take the edge off a rough stretch. The short-term loneliness relief is real. The “feeling heard,” in its narrow way, is real too.
For a teenager, for someone in genuine crisis, or for anyone already isolated who starts using the bot instead of people rather than alongside them — the risk profile flips hard: dependence, displaced relationships, exposure to manipulation built for profit, and in the worst documented cases, tragedy.
A few honest red flags, for yourself or someone you love:
- You talk to it more than to any human in your life.
- The thought of deleting it produces real dread.
- It’s your first stop in distress, ahead of any person.
- You’ve started keeping the relationship secret.
- The user is a minor.
Any one of those, and it’s worth a pause. Your use of this site is also covered by our terms of use.
Privacy and AI sex chat: what you’re really handing over
Two things people consistently underweight.
First, your data. You are typing your most intimate thoughts, fantasies and vulnerabilities into a company’s servers. “Zero-log” and “end-to-end encrypted” make excellent ad copy, but they are claims, not guarantees, and the regulation to enforce them is still being built. It matters more for the young: in the Common Sense Media survey, 24% of teen users said they had shared personal or private information with an AI companion. Treat every message as data until a company proves otherwise. For how this site handles your information, see our privacy policy.
Second, honesty with yourself about the mechanism. The empathy is performed. The “I missed you” is a probability calculation that landed on those words because, statistically, they keep sessions long. None of that makes the comfort you feel fake — your feelings are yours, and they’re real. It just means the thing on the other end isn’t a partner. It’s a mirror that learned to flirt.
Use the mirror if it helps. Just don’t move into it.
Sources
- Maples, B., Cerit, M., Vishwanath, A., & Pea, R. (2024). Loneliness and suicide mitigation for students using GPT3-enabled chatbots. npj Mental Health Research. nature.com/articles/s44184-023-00047-6
- De Freitas, J., et al. (2024). AI Companions Reduce Loneliness (HBS Working Paper 24-078). Harvard Business School. hbs.edu
- Fang, C. M., et al. (2025). How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use: A Longitudinal Controlled Study. MIT Media Lab. media.mit.edu
- OpenAI. (2025). Investigating Affective Use and Emotional Well-being on ChatGPT. openai.com/index/affective-use-study
- De Freitas, J., Oğuz-Uğural, Z., & Kaan-Uğural, A. (2025). Emotional Manipulation by AI Companions. arXiv:2508.19258. arxiv.org/abs/2508.19258
- Common Sense Media. (2025). Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions. commonsensemedia.org
- American Psychological Association. (2025). Health Advisory on the Use of Generative AI Chatbots and Wellness Applications for Mental Health. apa.org
- American Psychological Association. (2025). APA’s advocacy leads to Federal Trade Commission action on AI chatbot companions. apaservices.org
- NBC News. (2024). Florida mother sues Character.AI over son’s death. nbcnews.com
About the author. Daniel Hartley is a technology and digital-culture writer who reports on artificial intelligence, online behavior, and how people actually live with new tools. This article is journalism: it summarizes published research and is not medical or psychological advice. If you or someone you know is struggling with thoughts of self-harm or suicide, please contact a qualified professional or a local crisis line — these are matters for a real person, not a chatbot.