I Surveyed 2,150 Americans About AI Companions. Here’s What Surprised Me

Ranji Mercado — Wed, 09 Sep 2026 03:08:34 +0000

AI companions are usually discussed in extremes.

Depending on what you read, they're either harmless chatbots, a cure for loneliness, the future of relationships, or something that's going to make human connection even more complicated.

After spending months testing AI companion apps, I wanted something more useful than predictions. I wanted to know what people actually thought about them.

So in August 2026, I ran a U.S. survey through SurveyMonkey Audience asking adults about AI girlfriend and boyfriend apps, why they might use them, emotional attachment, privacy, and cheating.

We collected 2,150 responses. After a quality screen removed 441 responses showing a repeated first-option response pattern, 1,709 responses were retained for analysis.

Here are some of the findings that surprised me most.

1. About One in Four Respondents Had Already Used an AI Companion

Among the 1,709 retained respondents, 422 people, or 24.6%, said they currently used or had previously used an AI girlfriend or boyfriend app.

That was higher than I expected.

AI companions can still feel like a niche corner of the internet, especially when compared with general-purpose AI tools. But in this sample, roughly one in four respondents had firsthand experience with one.

And that experience turned out to matter.

2. AI Companion Users Were More Likely to Call AI Romance Cheating

This was probably the result I found most counterintuitive.

Among respondents who currently used an AI companion, 80.6% said a partner's romantic or sexual relationship with an AI could count as cheating.

Among former users, it was 60.2%.

Among people who had never used an AI companion, it fell to 44.0%.

I might have expected the opposite.

You could assume that someone who regularly talks to an AI girlfriend or boyfriend would be more inclined to say, "It's only a chatbot."

Instead, the people with firsthand experience were considerably more likely to recognize that an AI interaction could cross a relationship boundary.

That raises an interesting question: does actually using these systems make it easier to understand why an interaction with AI might feel meaningful enough to matter in a human relationship?

The survey can't answer why the difference exists, but the gap itself is difficult to ignore.

3. Men Were More Open to Developing Feelings for AI

We also asked respondents whether they thought they could develop real feelings for an AI girlfriend or boyfriend.

23% of men said yes, compared with 12% of women.

That's close to a two-to-one difference.

This is one of the areas I want to keep watching as AI companions improve. Emotional attachment isn't necessarily determined by how intelligent a model is. Things like personality, continuity, personalization, and memory can all influence how an interaction feels.

And memory, in particular, has stood out during my own testing.

An AI that remembers what you told it days ago feels very different from one that forgets everything as soon as the conversation moves on.

That's why I've also been running a separate AI companion memory test across popular companion apps.

4. Romance and Sex Weren't the Main Motivation for Most People

The phrase "AI girlfriend" naturally makes people think about romance and sex.

The responses were more complicated.

Among women, only 5.2% said affection, romance, or sexual interaction would be their main reason for using an AI companion.

Among men, the figure was 16.0%.

Men were roughly three times as likely to choose that motivation, but even among men it represented a minority of respondents.

Other possible motivations included having someone to talk to, feeling less alone, and feeling better or practicing social interaction.

That's important because it changes how we think about the category.

An AI companion can be romantic, but "AI companionship" and "AI romance" aren't necessarily the same thing.

The Hard Part Is Defining What an AI Companion Actually Is

Testing these apps has made me increasingly skeptical of treating AI companions as one single category.

Some interactions are essentially entertainment.

Some resemble role-playing.

Some people use them because they want someone available to talk to.

And some interactions can become persistent, personalized relationships in which the AI remembers details, develops a recognizable personality, and responds in ways designed to create emotional continuity.

That's where questions like "Is it cheating?" become much harder to answer.

The technology itself doesn't decide where that boundary is. People do.

And based on these responses, people don't agree on where the boundary should be.

Where I'm Publishing the Research

I've published the complete results and methodology on AI Girlfriend Coach.

I've also started maintaining an open AI companion research repository on GitHub where I'm documenting findings from the survey and other ongoing testing.

I'll continue adding research as I test more companion platforms and collect more data.

For me, the interesting question is becoming less "Can AI simulate a relationship?"

It clearly can.

The more interesting question is what happens when people start treating that simulated relationship as something that actually matters.