People Who Use AI Companions Take AI Relationships More Seriously

70.4% of users said a partner's AI romance could count as cheating, against 44.0% of never-users.

Current users were the strictest group in the whole survey, at 80.6%. Across all 1,709 respondents the figure was 50.5%, which is the headline number on the main survey page. The 422 users pull that average up; the never-users pull it down.

More revealing than the headline is which answer they chose.

40.5% of users said it becomes cheating once a partner grows emotionally or sexually attached, versus only 9.5% of never-users. Meanwhile just 14.0% of users said it cannot be cheating because an AI is not a real person, compared with 33.0% of people who had never tried one.

The biggest difference between the two groups was not whether users knew the other side was artificial. Of course they did. It was whether that fact made the relationship emotionally meaningless. Never-users were much more likely to say an AI relationship could not be cheating because the AI was not real. Users, meanwhile, were far more likely to locate the boundary at emotional or sexual attachment.

The four answers to the cheating question, and why the attachment answer matters so much, have their own post: does an AI relationship count as cheating?

74.2% of users could imagine developing genuine romantic feelings for an AI companion. Among never-users, 14.5% could.

More strikingly, 28.7% of users chose the firmest answer: they thought they could genuinely fall in love with an AI. Among never-users, just 1.6% did. What being open to those feelings does to the cheating answer, across the whole sample, is in the AI infidelity paradox.

65.4% of users could see themselves choosing an AI instead of pursuing a human relationship. Among never-users, 6.4% could.

This is the question I handle most carefully. It asked: if AI girlfriends and boyfriends became extremely realistic, could you see yourself choosing one instead of pursuing a human relationship? The figures above count only the two "yes" answers. That nearly tenfold difference was one of the widest gaps in the survey, far larger than the differences we saw between men and women or between generations.

Chatting, then feelings, then love, then a hypothetical choice over a person. At each step, users were far more open to the idea than people who had never tried an AI companion.

Former users consistently landed between current users and never-users, but much closer to the people still using the apps.

Three quarters of former users would consider using an AI companion again. Half could imagine choosing one over a human relationship.

Whatever separates users from non-users, deleting or abandoning the app does not appear to erase it.

A caution. We cannot tell whether using an app produced that difference or whether people with those attitudes were simply more likely to try AI companions in the first place. The survey shows the gradient and cannot say which way it runs.

Ask users the main reason for using an AI companion, and 56.2% choose having someone to talk to or feeling less lonely. Ask never-users, and 56.8% choose exactly the same thing.

The disagreement is in what comes second. 24.2% of users chose affection, romance or sexual interaction, more than four times the 5.7% among never-users. And only 6.6% of users chose coping with depression, against 22.4% of never-users.

People who had never used one were much more likely to picture AI companions as something for coping with depression; users were much more likely to choose affection or romance. Both groups agree the product is mostly about company, which fits what the loneliness numbers keep saying about who these apps are for. They disagree about what kind.

Asked what they would rather use if single, 58.3% of never-users chose dating apps alone. Among users, that fell to 33.6%.

But users did not simply switch sides: 37.4% said both, and 22.5% chose the AI outright. Among users in this sample, AI appears alongside dating apps as another option rather than simply replacing them. That squares with the dating app burnout numbers: people are tired of the apps, not necessarily done with them.

That is 64.5% of all 422 users in the sample.

I do not think the interesting interpretation is hypocrisy. It is that these respondents appear to grant AI relationships enough emotional weight for both statements to make sense at the same time.

That is not how people talk about software. It is how people talk about relationships. It is the same thing I kept running into when I looked at people who marry their AI: they know exactly what the machine is, and the attachment is real anyway.

These differences are associations, not evidence that using an AI companion changes people's attitudes. People who already saw AI relationships as meaningful may simply have been more likely to try the apps.

The 422 users also should not be treated as nationally representative of AI companion users. Reported use was substantially higher among Android respondents in our survey, even after quality screening. The attitude differences reported above, however, ran in the same direction across device groups.

Finally, 422 respondents gives us enough data for the user-versus-never-user comparisons on this page, but not for increasingly narrow cuts by age, gender or other characteristics. I have therefore left those out.

What the survey does show is a consistent association: people with firsthand experience of AI companion apps consistently treated AI relationships as more consequential than people who had never used one.

Every figure on this page traces to a table in the master research report. Users are defined as respondents who answered "Yes, I currently use one" or "Yes, I have used one in the past" to the question "Have you ever used an AI girlfriend or AI boyfriend app?" The rest of the method, including the 441-response screen, is on the survey's methodology section.