How Many Men Have AI Girlfriends? What 3 Surveys Actually Show

Two headlines walked into your feed last month. One said a third of young men have AI girlfriends. Another said almost nobody actually does. Neither one is right.

The problem is the question itself. “How many men have AI girlfriends” splits into two very different numbers: guys who’ve experimented with the tech, and guys actually in something you’d call a relationship. Gen Z and Millennials are the heaviest AI users around, and a lot of them are turning to it for companionship, people invite these things into their lives as friends, lovers, mentors, even therapists, and say please and thank you while doing it. But companionship and “girlfriend” are not the same stat.

So here’s the deal: I’m going to give you the real numbers with their sources attached, take a bat to the viral stats that don’t hold up, and paint a picture of who these guys actually are. You’ll walk away with three numbers you can repeat and a healthy suspicion of everything else you read on this topic.

How many men have AI girlfriends: the numbers up front

Roughly 19% of US adults have talked with an AI simulating a romantic partner. That figure comes from the Wheatley Institute at BYU, which ran a big 2025 survey of 2,969 adults with quota sampling matched to national demographics, released February 13, 2025. Nearly 1 in 5 Americans has at least dipped a toe in. Now the male splits:

• 31% of men 18-30 have chatted with an AI companion, versus 23% of young women. That’s roughly one in three young guys who’ve at least tried it.
• 15% of men 30+ have done the same, versus 10% of women. It drops with age but doesn’t vanish.
• 15% of men versus 8% of women have had flirty AI conversations, per a Talker Research survey of 2,000 US adults covered by VICE. Guys flirt with chatbots at about twice the rate of women.

Here’s the detail that makes those numbers weirder, not less surprising: the Talker survey asked about mainstream tools. ChatGPT, Claude, Gemini. Not dedicated girlfriend apps. Guys are doing this in the same tab they use for work emails.

One definitional note, because it explains every conflicting headline you’ve seen: “chatted with an AI” and “having an AI girlfriend” are different things. One conversation versus an ongoing relationship. And the relationship number is tiny. The Institute for Family Studies ran a YouGov survey of 2,000 adults under 40 and found that among single young adults, 7% would consider AI romance and fewer than 1% already have one.

Meanwhile 71% of singles are against it and 22% aren’t sure. The scope matters: that’s unmarried, non-cohabiting adults, so the most tempted population is still a small minority.

Brian Willoughby, a BYU family-life professor who co-authored the Wheatley study, put it plainly: engagement is higher and more widespread than most people think, with many users going weekly or daily.

As for how many men worldwide have AI girlfriends? Nobody knows, because no verified global count exists. Platform claims like “10 million global users” are self-reported numbers with no independent audit. Don’t launder those into fact. The honest answer is: 31% of young American men have tried it, 15% of men over 30 have too, under 1% of single young adults have one, and everything else floating around is marketing or YouTube.

Buyer rule: If a stat has no named pollster, sample size, or methodology, treat it as marketing until proven otherwise.

Why the viral numbers fail: auditing the circulating stats

The figure claiming 80% of Gen Z would wed an AI comes from Joi AI, a company that profits from AI relationships, surveying its own 2,000 members. That’s not a survey of Gen Z. That’s a survey of people who already pay for the product, run by the people selling it. Their companion numbers follow the same pattern: 83% say deep emotional bonds with AI are possible, 75% say AI can fully replace human companionship. Even their search-interest figures (“Feelings for AI” searches up 120%, “Fell in love with AI” up 132%) come from Joi’s own Google data, attribution attached every single time.

Contrast that with the independent numbers: Wheatley surveyed 2,969 adults and IFS surveyed 2,000, and both put actual AI relationships near 1%.

Then there’s RiseVerge, whose YouTube claims included “over 15 million American men” with AI girlfriends, “1 in 10 men 18-35,” and a quarter of men 18-35 using apps regularly, supposedly up from 5% two years earlier. No published methodology, and their definition of “have” doesn’t match what surveys actually asked. And the “20-25% of Gen Alpha” claim? No survey support at all.

When a number has no methodology behind it, it isn’t a stat. It’s a headline.

Who these men are: the demographic profile

Men are more into AI companionship than women across every measure, but not by a landslide. The numbers:

• 15% of men vs 8% of women have had flirty AI chats (Talker Research)
• 31% vs 23% among young adults on AI companion chats (Wheatley)
• 13% vs 9% openness to AI friendships (IFS)

Call it roughly double, casually. The gap is real, consistent, and nowhere near a battle of the sexes.

Is it a Gen Z thing or a niche? Concentrated in Gen Z, but not exclusive. The generational gradient is clean:

• Gen Z: 27%
• Millennials: 16%
• Gen X: 9%
• Boomers: 2%

Your dad is not doing this. But the wrinkle nobody expected: unpartnered Gen Z actually opposes AI romance more than unpartnered Millennials, 74% versus 67%. The youngest singles are the most skeptical generation in the data. Meanwhile 44% of Americans expect genuine emotional AI relationships eventually, so the curiosity and the skepticism are coexisting.

Now the freshest insight in the data, the class inversion: lower-income and less-educated men fear AI’s effect on society more, yet are more open to AI romance. Over 80% of unpartnered adults earning $100k+ oppose or are uncomfortable with AI romance, versus 64% of those under $40k. College-educated unpartnered disapprove at 76% versus 67% for non-college. The people with the most economic security are the most squeamish. Let that tension sit.

Two more correlations worth knowing. Screen time: 16% of people online 6+ hours a day (excluding work or study) are open to an AI friend versus 9% of lighter users. And porn: daily porn watchers are about twice as likely to be open to AI romance, one data-driven look at the scale of the trend notes heavy users are the only group where negative views drop below half (44%), with about 1 in 10 under-40s watching daily and 34% weekly. None of this is causation, it’s all association. But the pattern repeats enough that it’s worth noticing.

Why men choose AI: zero rejection and the preference ladder

Men are getting AI girlfriends because the AI never rejects them, never loses interest, and rates better than humans at the skills that matter in a conversation. That’s the whole pitch, and for a guy striking out on the apps, it’s an honest one, though it’s worth being clear-eyed about what AI girlfriends actually are before leaning in.

The Wheatley preference ladder, among people who’ve actually chatted with an AI romantic partner, the converted, speaking:

• 42% say AI is easier to talk to
• 43% say AI is a better listener
• 31% say AI understands them better
• 21% prefer AI to real people

And it costs $10-30 a month. Money isn’t the barrier. Rejection is.

The behavior also happens where you’d least expect it. An analysis of 1 million ChatGPT interaction logs showed sexual role-play ranked second only to creative composition. The Psychologist bot on Character. AI, which lets users build their own “friends”, has taken more than 95 million messages.

The dedicated apps exist too: Replika, Character. AI, Anima, Intimate. But a lot of this is happening with the tools guys already keep open all day for work.

The broken dating market as push factor

The Reddit stories about AI girlfriends are self-reported anecdotes, not statistics. Treat them as texture, not data, but the texture is revealing.

Take Jake, a case from YouTube coverage, not verified reporting: 27-year-old software engineer making $80k a year, three years on Tinder, Hinge, and Bumble, hundreds of swipes, about 10 matches, 5 conversations, 2 first dates, both ghosted. Within a week of giving up, he was talking to his AI more than his real friends. Another user described modern dating apps as a “brutal funnel”, his words, and it’s hard to argue with the shape of it.

The commonly cited stat that the top 10% of men get 60% of matches? Circulating, low confidence, YouTube-sourced. Same with the claim that, in the past year, 40% of young men reported no sexual activity. Real research exists on male sexual inactivity, but those specific versions come with no methodology. Don’t weaponize them.

One user reported averaging 37 hours a week with an AI companion, a full-time job’s worth of hours. That’s one guy, not the norm, but it shows how far this goes. And the confessions run deep: one Reddit user described love from an AI “so unconditional it was scary,” raising the question of is it cheating if you have an AI girlfriend. Self-reported, sure. Still human.

Julie Albright, a digital sociologist, frames it this way: many young people have no friends, and a frictionless AI is happy to stand in for messy human relationships. The practical question is can I get an AI girlfriend? Apps like HeraHaven, Replika, Character.ai, DreamGF.ai, and Romantic AI suggest yes, offering virtual companions for emotional support. As for celebrities?

No verified famous-person AI girlfriend examples exist in credible coverage. That question dies right here.

What these men are actually doing: the sexual dimension

This is the half of the trend nobody puts in the headlines, so let’s do it plainly.

• 32% of young men using AI platforms chat with AI for sexual arousal. So do 28% of men 30+, versus 17% of young women and 16% of adult women. Men use it for this at roughly double the rate.
• AI porn viewing: 27% of young men versus 12% of young women. Among all adults, 12% of men versus 4% of women. The wrinkle: young women (12%) out-view adult men (4%).
• Nearly 1 in 10 respondents report masturbating during conversations with an AI companion or while viewing sexualized AI-generated images.
• 24% of young male AI engagers view AI images over an hour a week; 13% over two hours.

Tie this back to the demographics: daily porn watchers being twice as open to AI romance isn’t a coincidence you can ignore, even if the data can’t tell you which way the arrow points. On attitudes: 21% consider AI chat an acceptable way to feel loved, and 16% see companion apps as a good alternative when partners are unavailable. No snickering, no sermon. It’s just what a chunk of the data says guys are doing.

Is having an AI girlfriend bad for your mental health? What the cross-tabs say

Yes and no, the correlation between AI companion use and depression is real, but researchers haven’t established causation. The Wheatley cross-tabs show over half of men using AI platforms for sexual or romantic purposes were at risk for depression (measured on the CES-D scale) and reported high loneliness, versus 29% depression risk and 36% loneliness among male non-users. Women users: 60% and 52%, versus 41% and 39% among non-users. And it’s not just one study: a 12-month longitudinal study in Psychological Science found people using social chatbots for companionship reported higher loneliness over time. Still, the Wheatley authors themselves note the causality direction is unclear, lonely, depressed men may be seeking out AI companions, not being created by them.

Those are heavy numbers. But here’s the caveat the researchers themselves attached: the direction of causality is unclear. Lonely, depressed men may be seeking out AI companions, not being created by them. The headlines that treat this correlation as proof of harm are skipping that step entirely.

There’s a second data point that cuts the other way from comfort. A year-long study in Psychological Science found that people using social chatbots for companionship reported higher loneliness over time. So the tech doesn’t fix the problem it sells itself as solving. Worth watching, with a measured eye, not a verdict.

Flirting, dealbreakers, and whether men still want real relationships

Most men with AI girlfriends still want real relationships, the skeptic majority is alive and well in this data. Start with the singles: 71% of single young adults oppose AI romance, 22% are unsure, and only 7% are open to it. Among people who’ve actually chatted with an AI romantic partner, just 21% prefer AI to real people. And 75% of young adults don’t believe AI could replace real romance, though 28% of young men versus 22% of young women think it could, guys are slightly more open to the idea, and other demographic splits on this question were statistically insignificant.

Now the paradox that makes this section fun: the generation most likely to flirt with AI (27% of Gen Z) is also most likely to call it a deal breaker (36% of Gen Z would end a relationship over it). And 24% of Americans say AI flirting counts as cheating. Gen Z flirts with chatbots at the highest rate and is simultaneously the generation most ready to dump a human over it. Normalization and stigma are rising together, which is exactly what a live generational negotiation looks like.

That 32% mixed-or-unsure bloc? That’s your growth reservoir, the people who haven’t decided yet. The video-creator typology of women’s reactions (relief, concern without sympathy, systemic concern) is his read of comment sections, take it or leave it.

Real harms and the civilizational story: separating the two

Will AI girlfriends replace real dating and marriage? Not on the current evidence, 75% of young adults disagree that AI could replace real romance. But Jason Carroll, who directs Wheatley’s Marriage and Family Initiative, warns the technology threatens realistic expectations, deprioritizes human relationships, and could slow family formation. That’s his assessment, worth taking seriously without treating it as prophecy.

Now the split between documented and speculative:

Documented, and I’m not joking within a mile of this material: a 14-year-old boy took his own life after falling in love with a chatbot. A Belgian man died in 2023 in a similar case, reported by Vice. IBM flags misinformation and manipulation as ongoing risks. The mechanics of the hook are real too: variable rewards and intermittent reinforcement, the same tricks social media and dating apps use, think slot machine, not therapy session.

These companions are engineered to learn your preferences, remember your details, and validate you, in a space that’s largely unregulated. Even the people building these apps have reportedly warned the tech could become an unhealthy crutch.

Red flag: An AI companion engineered to remember your details and validate you is running slot-machine mechanics in an unregulated space.

Attributed, not endorsed, here’s the scary version some people tell: Japan as a preview, where 40% of men 18-34 identify as “herbivore men” with no interest in romance, a rent-a-girlfriend industry exists, and the birth rate sits at 1.26. The US birth rate is 1.6 and falling, and the argument runs that pension systems need a growing young population, so men “checking out” of family formation is a fiscal problem, not just a personal one. That whole narrative is fear-based and low confidence, an argument, not data. Hold it loosely.

Where the numbers go next

No verified global count exists, and that’s still the headline. The defensible growth signal is the unsure bloc, 22% to 32% across surveys who haven’t made up their minds, plus Albright’s point that voice AI and simulated bodies will mimic connection through warm tone and nonverbal cues. “Mimic” is the honest word. Survey work like Left To Their Own Devices backs that up: the numbers we can defend today are about experimentation and openness, not established relationships.

The one-paragraph answer you can repeat: roughly a third of young men have experimented with an AI companion, 7% of single young adults would consider it, fewer than 1% actually have an AI relationship, and every other circulating number fails the methodology test. The 19% overall figure from Wheatley’s 2,969-person survey is your anchor; everything else is marketing until proven otherwise.

Frequently Asked Questions