The State of AI Romance in 2026: What 2,150 U.S. Adults Told Us

422 respondents had tried an AI girlfriend or boyfriend app, and exactly half were still using one.

That means 24.7% of the 1,709 analyzed respondents had first-hand experience with an AI girlfriend or boyfriend app: 12.3% were current users (211) and 12.3% were former users (211).

The perfect split gives us two equally sized groups who had tried an AI girlfriend or boyfriend app, but only one was still using it. What former users told us next makes that difference much more interesting.

74.4% of former users would consider using an AI girlfriend or boyfriend again.

For comparison, 96.2% of current users would consider future use. Among people who had never tried an AI girlfriend or boyfriend app, that number falls to just 11.7%.

Former users therefore looked much more like current users than never-users when it came to future interest. Stopping use didn't necessarily mean closing the door on AI companionship.

The survey can't tell us whether using an AI girlfriend or boyfriend app caused that openness or why former users stopped. But it does show a strong association between having experienced AI companionship and being willing to consider it again.

That difference between experienced and never-users will show up again later, when we ask whether a relationship with an AI can count as cheating.

The 422 AI girlfriend and boyfriend app users were almost evenly split between men and women.

Of current and former users, 50.7% were men (214) and 49.3% were women (208).

That doesn't mean men and women had identical usage rates. 13.5% of men were current users compared with 11.3% of women. But it does challenge the idea that actual experience with AI companionship is overwhelmingly male.

The more noticeable gender differences appeared not in who had tried one, but in how open men and women were to taking AI companionship further.

Men showed greater interest in the romantic side of AI companionship.

34.2% of men would consider using an AI girlfriend or boyfriend in the future, compared with 26.0% of women. Men were also more open to developing romantic feelings for AI: 34.5% versus 24.5%.

The largest gap appeared when respondents were asked why someone might use an AI companion. 16.0% of men chose affection, romance, or sex, compared with 5.2% of women.

That's roughly three times as high, but there's another way to read the same number: 84% of men still chose something other than romance or sex.

So the gender story isn't simply that AI companionship is "for men." The users in our survey were almost evenly split. Where men stood apart was in their greater openness to future use, romantic attachment, and the romantic or sexual appeal of AI companionship.

And even there, romance and sex were not the main draw. The next section shows what was.

56.6% chose having someone to talk to or feeling less alone.

That was by far the most common answer, compared with 18.5% for improving mental health or coping with depression, 14.6% for confidence or self-esteem, and just 10.3% for affection, romance, or sex.

In other words, respondents were more than five times as likely to choose company over romance or sex. Loneliness was also the leading answer across every subgroup we examined and across both major device groups.

Even among men, romance and sex remained a minority motivation.

Men were about three times as likely as women to choose affection, romance, or sex: 16.0% compared with 5.2%. But as we saw in the previous section, that still means 84% of men chose something else.

The broader pattern is therefore difficult to reduce to the stereotype of people seeking an artificial girlfriend or boyfriend for romance or sex. Respondents saw the appeal much more in having someone there.

This question was hypothetical and asked of all respondents, so it tells us what people believe might draw someone to an AI companion rather than the actual motivations of users.

But if companionship is the main perceived appeal, it raises a more complicated question: what happens when that companionship starts to feel emotionally real?

Half of respondents said AI romance can count as cheating.

50.5% (863 of 1,709) said it could be cheating. Another 21.2% wouldn't use the word cheating but would still be uncomfortable with it.

That means 71.7% placed some kind of relationship boundary around AI romance, while only 28.3% dismissed the issue on the basis that an AI isn't a real person.

Where respondents drew the line also mattered. 33.4% said any romantic or sexual AI interaction counts as cheating, while 17.1% said it only becomes cheating when emotional attachment develops.

For that latter group, the machine itself wasn't necessarily the problem. The attachment was.

The people closest to AI romance were actually the most likely to call it cheating.

Among people open to developing romantic feelings for an AI, 73.0% said AI romance could count as cheating, compared with 41.2% of those who said they could never develop those feelings.

The same pattern appeared with actual experience. 80.6% of current AI girlfriend or boyfriend app users called it cheating, compared with 60.2% of former users and 44.0% of never-users.

It's a counterintuitive result. The people most willing to take AI companionship seriously were not more likely to dismiss AI romance as harmless. They were more likely to treat it as something consequential enough to become a betrayal.

The survey can't tell us whether experience with AI caused those attitudes. But it does suggest that the more emotionally plausible AI relationships become to someone, the harder they may be to dismiss as "just software."

Engaged or married respondents were nearly twice as likely as singles to have used an AI girlfriend or boyfriend app.

30.7% of engaged or married respondents had used an AI girlfriend or boyfriend app, compared with 16.3% of single respondents.

That's almost a two-to-one difference. It challenges the assumption that AI companions are mainly an alternative for people who don't already have a partner. In our sample, experience with AI companionship was actually more common among people who were engaged or married.

But being more likely to have used an AI girlfriend or boyfriend app didn't make them more permissive about AI romance.

54.8% of engaged or married respondents said AI romance can be cheating, compared with 42.2% of singles.

That's a 12.6 percentage-point gap, and importantly, the pattern remained consistent across device groups.

Put the two findings together and an interesting tension appears. Engaged and married respondents were nearly twice as likely as singles to have used an AI girlfriend or boyfriend app, yet they were also more likely to say a romantic or sexual relationship with one could cross the line into cheating.

Having experience with AI companionship, then, doesn't necessarily mean dismissing AI intimacy as harmless or insignificant.

Engaged or married women drew the strictest boundary around AI romance.

Among engaged or married women, 39.8% said any romantic or sexual interaction with an AI already counts as cheating, compared with 30.1% of engaged or married men and 21.9% of single men.

That's stricter than saying cheating begins only when emotional attachment develops. For this group, the interaction itself was enough to cross the line.

44.7% of Millennials would consider using an AI girlfriend or boyfriend, the highest of any generation.

Future interest varied sharply by age:

• 44.7% of Millennials
• 31.6% of Gen X
• 12.5% of Gen Z
• 8.6% of Boomers

Millennials were more than three times as likely as Gen Z to consider an AI girlfriend or boyfriend. Gen X was also more than twice as likely, and that direction remained consistent across device groups.

Gen Z was much closer to Boomers than to Millennials.

Only 3.9 percentage points separated Gen Z's 12.5% from Boomers' 8.6%. The gap between Gen Z and Millennials, meanwhile, was 32.2 points.

The survey can't tell us why this pattern exists, and we shouldn't assume that age itself explains it. But it does challenge a simple younger-means-more-interested narrative. In our sample, the strongest interest in AI companionship came from the middle generations, not the youngest adults.

Of course, considering an AI girlfriend or boyfriend doesn't necessarily mean choosing one instead of another person. Our final set of findings makes that distinction particularly clear.

Dating apps were preferred over an AI girlfriend or boyfriend by nearly 7 to 1.

When asked which they would prefer for finding companionship or a relationship, 52.2% chose dating apps, compared with just 7.8% who chose an AI girlfriend or boyfriend.

Another 13.2% would use both, while 26.8% were unsure. Even among singles, only 6.1% (19 of 313) chose an AI girlfriend or boyfriend.

64.6% said AI could never replace a human relationship in their life.

This result comes with a device caveat: the figure was 50.2% among Android respondents and 79.1% among iOS and desktop respondents. Still, it adds an important distinction to the rest of the research.

Across the survey, 29.9% said they would consider an AI girlfriend or boyfriend, yet only 7.8% preferred one over dating apps. These aren't contradictory findings. Being open to AI companionship isn't the same as wanting AI instead of another person.

And that may be the bigger story running through the research. People can believe AI will never replace human relationships while still treating attachment to AI as meaningful enough to have boundaries around it. AI companions don't need to replace human relationships to start affecting the rules of human relationships.