Get the app
How it worksThe castCompareGuidesPricingFAQ Languages About

Guide

Who actually uses AI companions? The numbers that exist, and the ones that don't

Five independent surveys since 2025 put usage anywhere from 4% to 72%, because they ask different questions of different people. No audited count of AI companion users exists.

By LORAX Team Published Updated

Five independent surveys have measured this since early 2025, and their headline figures run from 4% to 72%. They disagree because they ask different questions of different people over different time windows. No audited count of AI companion users exists anywhere.

That second fact matters more than the first. Everything below is either a survey with a published base, or a number a company says about itself, and the two are labeled separately.

The independent surveys, with their bases attached

Source and date Who was asked Question Figure
Wheatley Institute, Feb 2025 2,969 US adults Ever chatted with an AI meant to simulate a romantic partner 19%
Common Sense Media, Jul 2025 1,060 US teenagers Ever used an AI companion 72%, with 52% at least a few times a month
Gallup for the Walton Family Foundation, reported Jan 2026 around 2,500 US young adults Used an AI chatbot as a girlfriend or boyfriend in the last month 10%
Wheatley Institute and Institute for Family Studies, May 2026 2,431 US young adults in committed relationships Regularly interact with an AI chatbot simulating a romantic partner 15%, with a further 20% to 30% having experimented
Pew Research Center, Jun 2026 5,119 US adults Use AI chatbots for companionship 4% of chatbot users, and 10% for emotional support or advice

All five are independent of any companion app company, and each names the organization that ran it and the size of its sample. The Counterfeit Connections team also published a peer-reviewed article on the same question in the Journal of Social and Personal Relationships in 2025.

Why the numbers disagree by a factor of five

Three things move the figure, and none of them is measurement error.

The first is the recall window. “Ever chatted” collects everyone who tried it once out of curiosity. “In the last month” collects people who are actually doing it. That alone explains most of the distance between the Wheatley figure of 19% and the Gallup figure of 10%.

The second is who counts as a companion. Pew asked about AI chatbots in general and then asked what people use them for, which is why its companionship number is 4%. The relationship surveys asked directly about a chatbot that simulates a romantic partner. Different populations, both correctly measured.

The third is the age band. Every survey that isolates young adults finds a higher rate than the same survey’s figure for all adults.

Men and women use them at rates closer than the stereotype

The February 2025 survey found 31% of young adult men and 23% of young adult women had chatted with an AI boyfriend or girlfriend. Among older adults the split was 15% of men and 10% of women (Wheatley Institute, 2025).

The 2026 follow-up reported men somewhat more likely to use one regularly, and described the gender difference as generally small (Wheatley Institute and Institute for Family Studies, 2026).

The category is not a men’s category. It is a men-somewhat-ahead category, which is a different claim and the only one the data supports.

Teenagers are measured differently from adults, so treat the 72% carefully

Common Sense Media’s July 2025 survey defined AI companions broadly for its respondents, as digital friends or characters you can talk to. On that definition 72% of US teenagers had used one and 52% used one at least a few times a month. Thirty-three percent used them for social interaction and relationships, a category that includes practicing conversation, emotional support, role-play and friendship as well as romance (Common Sense Media, 2025).

That 72% is not 72% of teenagers with an AI girlfriend, and it gets quoted as though it were. The same survey found 46% of teenagers describe these things as tools or programs rather than friends.

What the vendor numbers are worth

Vendor-reported figures are not worthless. They are just a different kind of claim, and they should never be mixed into a table with survey data.

Replika’s homepage carried a live counter reading 42,160,934 users worldwide when we opened it on 13 September 2026. There is no published definition of a user attached to it, no statement of whether it means registered accounts or active people, and no independent audit.

Character.AI’s own about page says millions of people visit every month, and gives no number at all.

Market-size and total-user estimates for this category circulate widely. We found none that published a sample, a question and a date together, so none is cited here. If a figure has no sample, no question wording and no date, it is not a statistic.

We make LORAX, an AI companion app for iPhone; we don’t publish its user numbers, and none of the figures on this page are ours.

That is a deliberate position rather than modesty. Self-reported user counts in this category cannot be checked by a reader, so publishing ours would add a number to the internet without adding any information to it.

What nobody has measured yet

Four gaps, and they are the reason any confident total you see online is made up.

There is no longitudinal series. Every figure above is a snapshot, so nobody can say whether use is rising, flat or already falling, only that different surveys in different months produced different numbers.

There is almost no data outside the United States. All five surveys here are US samples, in a product category sold worldwide.

Nobody has measured retention. Not one published survey reports how long people keep using a companion app, which is the single number that would tell you whether this is a habit or a curiosity.

And the definitions have not settled. Until surveys agree on what an AI girlfriend actually is, the figures will keep disagreeing for reasons that have nothing to do with behavior.

For what the same research says about effects rather than headcount, what the research says about harm covers both directions. For the secrecy data behind the prevalence figures, is an AI girlfriend cheating goes through it, and is it weird to have an AI girlfriend takes on the question most people are really asking.

Sources

All links checked 2026-09-13.

Questions people ask

How many people use AI companions?

It depends entirely on the question asked. A February 2025 survey of 2,969 US adults found 19% had ever chatted with an AI romantic partner. A Gallup survey reported in January 2026 found 10% of young adults had used one in the last month. Pew found 4% of US chatbot users use chatbots for companionship.

Who uses AI girlfriends, men or women?

Both, with men somewhat ahead. The Wheatley Institute survey found 31% of young adult men and 23% of young adult women had chatted with an AI boyfriend or girlfriend. Its 2026 follow-up found men slightly more likely to use one regularly, with a small gender gap.

Are these AI girlfriend statistics reliable?

The survey figures come from named organizations with published sample sizes and methods, so they can be checked. The market-size and total-user figures circulating online mostly cannot, because the firms selling them do not publish how they were produced.

How many users does Replika have?

Replika's own homepage showed a counter reading 42,160,934 users worldwide when we checked it on 13 September 2026. That is a vendor-reported figure with no published definition of a user and no independent audit.

Do you publish LORAX user numbers?

No. We do not publish user counts, download counts, ratings or revenue for our own app, and none of the figures on this page are ours.

Meet someone who texts first.

LORAX is free to download and free to talk. Credits and Premium are optional and always priced before you spend.

iPhone · free · no account needed to look around