0. Bottom line: “the people behind the avatars are both men” only explains half the experience
A Japanese study surveyed 217 VRChat users aged 18 to 39 and found a pattern that looks strange if you classify everything only by the sex of the person behind the avatar.
Among 93 respondents categorized by the survey as biologically male who had experienced virtual sex, 54 said their past partners had been male only. Just seven said female only. Even among 46 men who identified as heterosexual, only three reported female-only partners.
That immediately invites the joke: “So it is just men with men?”
At the level of the people behind the avatars, male-male pairings were indeed common.
But that is not necessarily what the user is seeing.
The user sees an avatar. The other person appears as an avatar. The user may also see their own avatar in a mirror.
That suggests a different way to think about it:
VRChat virtual sex may sometimes feel less like a direct copy of real-world sex and more like a multiplayer, first-person version of a 3D character game.
This is an interpretation, not a conclusion proven by the study. But it makes the data much easier to understand.
1. The numbers are already unusual
The survey was conducted from November 15 to December 15, 2024, among 217 VRChat users living in Japan who used Japanese in daily life. Participants were recruited through X.
Key figures:
| Measure | Result |
|---|---|
| Total respondents | 217 |
| Had experienced virtual sex in VRChat | 103 (47.5%) |
| Survey-category male respondents with experience | 93 of 176 (52.8%) |
| Survey-category female respondents with experience | 10 of 40 (25.0%) |
| Experienced male respondents: male-only partners | 54 of 93 (58.1%) |
| Experienced male respondents: female-only partners | 7 of 93 (7.5%) |
| Experienced male respondents: included a male partner at least once | 79 of 93 (84.9%, calculated from Table 5) |
| Heterosexual experienced men: female-only partners | 3 of 46 (6.5%) |
| Heterosexual experienced men: included a male partner at least once | 38 of 46 (82.6%, calculated from Table 5) |
| Survey-category men using a human female avatar | 138 of 176 (78.4%) |
Among the 138 male respondents using human female avatars, only 35 clearly answered that they were “playing a sex different from the real world.” Another 42 answered that it was hard to say.
So even “male user + female avatar” does not automatically mean “a man role-playing a woman.”
The more carefully you read the tables, the less useful a one-line real-world label becomes.
2. The study does not tell us whether the visible scene was “two girls”
This limitation matters.
The survey recorded the perceived biological sex of past partners. It did not record:
- the respondent’s avatar appearance during each encounter,
- the partner’s avatar appearance during each encounter,
- whether the user was mainly in first-person view,
- whether they watched both avatars in a mirror,
- what exactly they found arousing or appealing,
- whether the main attraction was character design, intimacy, role-play, visuals, or something else,
- or how often each kind of encounter occurred.
So the paper cannot tell us how often the following happened:
people behind the avatars: man + man
visible scene: female avatar + female avatar
It also cannot prove that users were deliberately creating a girls-love or “yuri” scene.
That remains a hypothesis.
3. Think of 3D character culture and the behavior becomes less mysterious
3D character games and animation tools have long supported a familiar style of play:
make a character, pose characters, choose expressions, arrange a scene, and enjoy the final composition.
Koikatsu offered detailed 3D character creation, scene creation through Chara Studio, and VR features. Custom Order Maid 3D2 also includes a studio mode where characters and items can be placed freely to build a scene.
In this kind of play, the user can be several things at once:
character designer
operator
director
viewer
The experience is not necessarily pure self-insertion.
VRChat adds one major twist: you can step inside the scene.
4. VRChat turns it into first-person multiplayer
In a conventional studio mode, the user usually stands outside the scene and controls it.
In VRChat, the camera effectively moves into the character.
Now the user can:
- see the other avatar directly in front of them,
- move and have their avatar move with them,
- use a mirror and see both avatars together,
- interact with another avatar controlled by another human.
That changes the structure from:
“I arrange two 3D characters and watch them”
to:
“I inhabit one of the 3D characters, while another human inhabits the other one.”
That is why “a first-person multiplayer 3D dating sim” is a surprisingly useful analogy.
5. The biggest difference is that the other person is not an AI or a fixed animation
The other side is human.
That means they can react.
If you move closer, they can respond. If you stop, they can wait. If you joke, they can joke back. If something feels wrong, either person can change what they are doing.
The scene is being built together in real time.
That is very different from simply watching a prepared animation.
A prior survey cited in the paper explicitly framed VR sexual activity as interpersonal sexual communication through VR, not a computer-controlled game partner.
Visually, it may resemble 3D character culture.
Socially, it is still person-to-person interaction.
That combination is the important part.
6. The sex of the person behind the avatar and the visible scene can diverge
In physical life, the other person’s body and the visual stimulus are usually the same object.
In VR, several layers can separate:
- Who you believe the other person is.
- What their avatar looks like.
- What your own avatar looks like.
- What kind of story or scene you think you are in.
- What your relationship with the other person means to you.
So the people behind the avatars may both be men while the visible scene could involve two female avatars.
And the user’s subjective experience does not have to be “I am literally a woman now.”
It could be:
“I am operating this character.”
“I like seeing these two characters together.”
“I like making this scene happen.”
“I mainly like the other person responding to me.”
Several of these can be true at the same time.
The study’s finding that self-identified heterosexual men frequently reported male partners suggests that real-world sexual-orientation labels do not perfectly predict VR behavior.
It does not mean one VR encounter reveals someone’s “true” orientation.
7. Four different kinds of enjoyment may overlap
The study did not directly measure these motives, but the medium makes at least four modes possible.
1. Embodying the avatar
The avatar starts to feel like the body through which you are acting.
2. Playing a character
You behave as a character rather than simply as your offline self.
3. Building and viewing a scene
You enjoy the visual pairing and composition, similar to character-scene creation.
4. Enjoying live interaction
The core appeal is that another person reacts intelligently and unpredictably.
These modes are not mutually exclusive.
VRChat can let someone embody a character, perform a role, enjoy the scene, and improvise with another human at the same time.
No wonder a simple category chart starts to look confused.
8. So is it “men with men,” “two girls,” a game, or a relationship?
The answer depends on which layer you are measuring.
If you count the people behind the avatars, male-male pairings were common in this sample.
If you look at avatars, the visible pairing may be completely different.
If you ask about subjective experience, one person may feel embodied, another may feel like a director, and another may mainly care about the relationship.
Trying to force one label onto all of those layers creates the confusion.
The most useful current interpretation is:
VRChat virtual sex may be easier to understand as a first-person, multiplayer extension of 3D character culture than as a simple digital copy of real-world sexual behavior.
But the next study needs to connect the missing pieces: the user’s avatar, the partner’s avatar, viewpoint, motive, and what part of the experience actually mattered.
Only then can we properly answer why so many self-identified heterosexual men in this sample reported male partners.
References (4)
Checked: October 6, 2026
- Digital Games Research Association Japan, “The situation of virtual sex among young users in VRChat”
https://www.jstage.jst.go.jp/article/digrajproc/15/0/15_139/_article/-char/ja/ - PDF
https://www.jstage.jst.go.jp/article/digrajproc/15/0/15_139/_pdf/-char/ja - Preserved Koikatsu concept page
https://old.illusion.rip/preview/koikatu/concept/index.php - Custom Order Maid 3D2 system page
https://com3d2.jp/com3d2/system.html
