People Who Believe They Could Love an AI Are More Likely to Call AI Romance Cheating

For this analysis I combine the first two groups as "open to romantic feelings for an AI." That gives 500 respondents.

Most of them, 358 of the 500, did not say they could fall in love with an AI. They said they might, under a condition the technology has not met yet. Nobody in this group was asked whether they are in love with an AI now, and the question says nothing about whether an AI can feel anything back. It is entirely about what the human respondent could imagine feeling.

So "believers" means people who did not rule it out. That is a lower bar than the word suggests, and it makes the finding stronger, not weaker.

A 31.8-point gap, in the direction most people would not have predicted.

If you stopped here, the natural conclusion would be that believers are simply stricter about everything to do with AI and relationships.

They are not. The individual answers show something more specific.

Look only at the harshest answer, that any romantic or sexual interaction with an AI is cheating. Among people open to AI feelings, 32.4% chose it. Among people who said never, 33.8% did.

Almost identical. The 31.8-point gap in the combined figure is not coming from believers deciding that a partner talking romantically to an AI is automatically an affair. On that question, the two groups agree.

The gap is coming from somewhere else entirely.

Among respondents open to AI feelings, 40.6% chose it. Among respondents who said they could never develop such feelings, 7.4% did.

That is where the whole paradox lives. The strictest answer is flat across the two groups. The uncomfortable-but-not-cheating answer is flat too, at 19.8% versus 21.7%. The one answer that moves is the conditional one, and it moves by 33 points.

One possible interpretation is that people who can imagine AI producing genuine romantic feelings are more likely to treat emotional attachment itself as consequential, regardless of whether the other participant is human. The survey did not ask respondents why they chose the answer they did, so this is a reading of the pattern rather than a reason they gave. But it is the reading the numbers most obviously support: for believers, the line sits not at the interaction but at the attachment.

Among people open to AI feelings, 7.2% said a partner's AI romance is not cheating because an AI is not a real person. Among people who said they could never develop AI feelings, 37.1% gave that answer.

More than five times the share. The two moving answers are almost mirror images of each other, and together they give the article its clearest contrast. If you do not believe genuine feelings toward an AI are possible, "it's not a real person" can settle the cheating question on its own, and for more than a third of that group it did. If you believe those feelings could become real, the humanity of the AI appears to matter less than the attachment of the human. Again, that second sentence is an interpretation of what respondents chose, not something they told us in those words.

Written as a headline, the finding looks like a contradiction: open to AI love, strict about AI cheating.

Written as the individual answers, it looks internally consistent.

Someone can believe that humans can develop genuine romantic feelings toward an AI. They can believe those feelings can become emotionally significant. And from those two beliefs it follows that a partner developing such feelings toward an AI could violate a relationship boundary. The same premise supports both positions.

Seen that way, the surprising group is not the believers. It is the skeptics, whose answer to the cheating question depends on a claim about the AI rather than about their partner. The believers are answering a question about a person. 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 to them anyway.

The caveat, kept visible. The survey shows an association between two answers. It does not show that believing in AI love causes anyone to view AI romance as cheating, and the master report is explicit that relationships between variables in this survey are associations, not causes. Both answers could be driven by something the survey did not measure.

Among respondents who currently use an AI girlfriend or boyfriend app, 80.6% (170 of 211) said a partner's AI romance can count as cheating, and their most common answer was the attachment one.

The pattern is not confined to the feelings question. People with direct experience of these products were, if anything, even less inclined than the believers to say the question does not apply.

That group deserves its own treatment and has one, in our analysis of people who have actually used AI companions. For this article the relevant point is narrower: two different ways of being close to AI relationships, imagining feelings and actually using the apps, both go with taking AI infidelity more seriously.

The survey measures attitudes. It does not establish what infidelity objectively is, and nothing here should be read as 73% proving that AI relationships are cheating.

It also does not tell us whether respondents had personally experienced AI-related conflict, whether they were in monogamous relationships at the time, what boundaries they had already agreed with a partner, whether their answer would change for sexting versus roleplay versus emotional disclosure versus secrecy, or why they chose the answer they did. The wider results on how respondents overall answered the cheating question are in our analysis of whether an AI relationship counts as cheating. This page is about one thing: who is most likely to perceive it that way, and the fact that it is the people you would least expect.