Who Uses AI Girlfriends? Demographics and Statistics

Updated 2026-06-10. AI girlfriend guides

Some of it is the stereotype people expect: early adoption skewed young and male. But that picture is at least three years out of date, and the specific numbers attached to it are not sourced.

This page previously reported a median user age of 27, a 75/20/5 gender split, exact percentages per age bracket, per-country user counts including "9 million" in the US and "4 million" in Brazil, and psychographic segments broken down to the decimal. None of it can be verified. There is no industry-wide demographic survey of AI companion users. What exists is a handful of small, self-selected academic studies — and they describe very different populations.

Correction, October 5, 2026. Every percentage and user count previously on this page has been removed. The claim that over-45 users show "the highest retention and satisfaction rates of any age group" was also unsourced and is gone.

What the real studies say

China, 612 long-time users (Liu, Frontiers in Psychology, 2025): a large survey of long-term users of AI companion apps, which is more representative of committed users than of the general user base. Because it sampled from people already using these apps long-term, its numbers describe a more engaged population than average.

A 12-month study of 1,182 Character.AI users (arXiv 2609.07243): a large two-wave longitudinal sample, with 439 responding at roughly 12 months. Character.AI's audience skews younger and more male than the overall category, so it is not representative either — but it is one of the larger real datasets in existence.

A preregistered 21-day randomized trial of 183 participants (arXiv 2509.19515): the only randomized evidence we could find, with participants assigned to chat with Replika or play word games as an active control. It found no significant difference in social health or loneliness, and found that people who treated the chatbot as more human-like reported larger effects on their human relationships.

Three studies, three different populations, none of them a demographic census. That is the honest state of the field.

What can be said with reasonable confidence

Adoption skews young. Every dataset that exists points in this direction. The exact brackets are not knowable, but 18-34 is clearly the core segment.

The user base skews male, though the gap is narrowing. Early romantic-companion products were heavily male, and the gap persists, but the direction of travel is toward balance. Platform-level variation is significant — products with broader use cases skew less male than those selling specifically a romantic companion.

Use extends well beyond the lonely and single. A meaningful minority of users are in relationships, and the reasons vary from shared entertainment to private exploration. This group is genuinely under-studied and the ethics of it are unresolved.

Non-Western adoption is substantial, not peripheral. Southeast Asia, Latin America, and India all have significant user bases. Cultural attitudes toward dating and appropriate social outlets differ substantially across regions, and the reasons people reach for a companion are not uniform.

What cannot be said

Exact age distribution. Gender ratios. Per-country user counts. Psychographic segment shares. Retention or satisfaction rates by demographic.

If you need these for research, your options are limited: published academic studies, which each describe one sample, or your own product analytics, which describe your users. There is no third source, and the numbers circulating online are almost all restatements of each other without primary provenance.

Why we removed the numbers

We are aware this makes the page less useful-looking. But a demographic breakdown presented with four significant figures implies a precision that nobody possesses, and several of the figures we removed would not survive a citation request. If you are doing real research on who uses AI companions, working from the three studies above will serve you better than the invented table did.

Frequently Asked Questions

What percentage of AI girlfriend users are women?

Nobody has published a defensible figure. Romantic-companion platforms skew heavily male, and the gap is narrowing, but the specific percentages commonly cited have no traceable source.

What is the average age of AI girlfriend users?

There is no reliable average. The available datasets skew young, but they sample from different populations — a Replika trial, a Character.AI study, and a Chinese long-term-user survey would give you three different answers, all valid for their own sample.

How many users are in relationships?

Unknown in aggregate. This subgroup is discussed in the ethics literature as a category but has not been measured. Which is unfortunate, given it is probably the most interesting question in the field.

Where do most users live?

There is no reliable breakdown. The US has the largest absolute user base of any country we can point to, and Southeast Asia shows high per-capita adoption — but specific country counts circulating online are estimates built on download data that misses web usage entirely.

This article is for informational purposes only and does not constitute medical or professional advice. Luvlock AI provides tools for companionship but encourages maintaining a balanced approach to relationships and mental health.

Ready to experience AI companionship for yourself? Get started with Luvlock AI →