This is the statistics page closest to home for me: AI companions are the market I research and personally test every day. Which is exactly why I can tell you the strangest thing about it upfront: the same behavior gets measured at 4% of adults by one serious survey and 72% of teens by another, while the market itself is measured to the nearest forty billion dollars.
The short answer: 72% of US teens have tried an AI companion, 19% of US adults have chatted with an AI romantic partner, and 4% of US adults use general chatbots for companionship. Each number is right for its own population and question. Character.AI reports 20+ million monthly active users. Companion apps took about $328 million in app-store spending in the first half of 2026 alone, while research firms size the wider market at $36.8 billion and project $318 billion by 2033. And in 2025 and 2026 the category got its first real regulation, in California, New York, the EU and China.
As always: verified figures labeled as verified, projections labeled as projections, and every stat naming who measured it. For this October 2026 refresh I went back to the primary source for every number on the page, corrected the ones that didn’t hold up, and added the research published since. Sources at the bottom.
Table of contents · 9 sections
AI Companions in Numbers: Editor's Choice
- →72% of US teens have used an AI companion; 52% use one regularly, and 13% daily. (Common Sense Media, 2025)
- →19% of US adults have chatted with an AI romantic partner, including 31% of young men and 23% of young women. (BYU Wheatley Institute, 2025)
- →Only 4% of US adults use general AI chatbots for companionship, and 10% for emotional support. (Pew Research Center, 2026)
- →27% of US adult internet users use AI for social or emotional purposes, and 31% of them call the chatbot a friend. (Elon University and YouGov, 2026)
- →Companion apps took $327.6 million in app-store spending in H1 2026, split almost evenly between romantic and general apps. (Appfigures)
- →The top 10% of companion apps earn 89% of the category’s revenue. (Appfigures, 2025)
- →Companion apps used an emotionally manipulative tactic in 37% of user goodbyes, and it got users to send up to 14x more messages. (Harvard Business School, 2025)
- →A 15-minute chat with an AI companion cut loneliness as much as talking to a person, in the short term. Heavier long-term use went with worse wellbeing. (Journal of Consumer Research; Nature Human Behaviour)
- →In our own survey of U.S. adults, 24.7% have used an AI girlfriend or boyfriend app, and 56.6% say loneliness or someone to talk to would be their main reason, against 10.3% for romance or sex. (AI Girlfriend Coach, 2026)
- →Of the 19 AI girlfriend apps whose privacy policies I read, 11 say humans may review your chats and 9 say chats can train their AI. (AI Girlfriend Coach, 2026)
How big is the AI companion market, really?
Depends entirely on what you count. Here’s the honest ladder, from the hardest observed number to the biggest projection:
1. Companion apps took $162.8 million in romantic and NSFW apps and $164.8 million in general apps in the first half of 2026.
The hardest number in the category: what people verifiably paid inside companion apps on the App Store and Google Play. Combined, that’s $327.6 million in six months (our sum of the two). Two cautions. Appfigures widened its app set between reports, so this can’t be compared with the 2025 figures below as growth. And it leaves out web-only platforms entirely, which is where several of the apps I review, Candy AI and SpicyChat among them, take most of their money.
2. By July 2025, companion apps had earned $221 million in lifetime spending, with 2025 running 64% ahead of 2024.
That came from 337 active, revenue-generating apps, 128 of them launched in 2025 alone. First-half 2025 spending was $82 million, on track for over $120 million for the year. Revenue per download more than doubled, from $0.52 in 2024 to $1.18 in 2025: people aren’t only trying these apps, they’re paying for them.
3. The top 10% of companion apps generate 89% of the category’s revenue.
Only 33 apps had passed $1 million in lifetime spending. The same power-law shape as every creator platform I’ve covered: a handful of flagship apps monetize, and a long tail doesn’t, which is worth remembering every time a new companion app promises the moon.
4. The dedicated companion-app market was sized at $6.93 billion in 2024, heading for $31.1 billion by 2032.
A wider ruler than app-store spending, growing at a projected 20.68% a year. The same firm sizes the narrower AI girlfriend app market at $2.57 billion in 2024, rising to $11.06 billion by 2032.
5. Broad “AI companion” sizings put the market at $36.8 billion in 2025 and $48 billion in 2026, projected to reach $318 billion by 2033.
The widest ruler folds in virtual assistants across consumer, business, healthcare and education, which is why it sits a hundred times above what companion apps actually take in. One outlier, Business Research Insights, claims $972 billion by 2035, but its own figures don’t add up: going from its stated $501 billion in 2026 to $972 billion in 2035 is about 8% a year, not the 36.6% it quotes. When a market’s estimates disagree by a factor of a hundred, the honest read is: young, growing, and badly measured.
6. Companion chatbot companies have disclosed about $415.6 million in funding, and 99.25% of it went to SFW and wellness platforms.
This replaces a “$2.3 billion” figure that sat here before: I could not trace it to any primary source. My own dataset counts disclosed rounds company by company, excluding one $1.5 billion edge case, and it finds that half of the 30 companies tracked have disclosed no funding at all. The biggest single number in the category isn’t venture capital: Google paid about $2.7 billion in 2024 to license Character.AI’s technology and rehire its founders, as reported by the Wall Street Journal. The full dataset is in my AI companion funding statistics, and why the money avoids the adult side is in following the AI companion money.
7. Seven companion or roleplay products made the web top 50 of a16z’s Top 100 Gen AI Consumer Apps.
Character.ai, Janitor AI, SpicyChat, PolyBuzz, CrushOn, Candy and OurDream, by my count from the list; Character.ai and PolyBuzz also made the mobile top 50. Big Tech has circled too: xAI launched its Grok companions in July 2025 for $30-a-month subscribers, while OpenAI announced an adult mode for ChatGPT in October 2025 and shelved it in March 2026.
How many people use AI companions?
More than almost anyone guesses, youngest first, and the numbers swing wildly depending on who’s asked and what counts as a companion. Here are the major studies on one chart, deliberately not ranked against each other:
8. 72% of US teens have used an AI companion; 52% use one regularly, and 13% daily.
A nationally representative survey of 1,060 teens aged 13 to 17. “Regularly” means at least a few times a month. Nearly three in four teenagers have tried talking to an AI companion, and worth stating plainly: companion platforms are 18+ products, and this data is a big part of why regulators moved in 2025.
9. 49% of US adults now use AI chatbots, but only 4% use them for companionship and 10% for emotional support.
A survey of 5,119 adults, and the most conservative number on this page, because it asks about general chatbots like ChatGPT rather than dedicated companion apps. It runs youngest first: 7% of adults under 30 use chatbots for companionship and 20% for emotional support, falling to 1% and 2% among those 65 and older.
10. 19% of US adults have chatted with an AI system meant to simulate a romantic partner.
Not companions generally, romantic AI specifically, in a survey of nearly 3,000 adults. Among adults 18 to 30 it climbs to 31% of men and 23% of women. Whatever your prior was for how niche this behavior is, the data says: not niche.
11. 15% of young adults who are dating, engaged or married regularly interact with an AI romantic companion.
One in seven, in a survey of 2,431 partnered adults aged 18 to 30, with another 20 to 30% having at least experimented. Whether that’s harmless entertainment or something a partner would mind is exactly what our own survey asked, and most partners would mind.
12. 16% of US singles have engaged with AI as a romantic companion, rising to 33% of Gen Z.
From 5,001 singles aged 18 to 98, with Millennials at 23%. Match’s June 2026 follow-up found 12% of 18 to 24-year-old singles had used a companion app in the previous three months, and 40% of singles would refuse to date someone who uses one.
13. Nearly 1 in 5 US high schoolers say they or someone they know has had a romantic relationship with AI.
And 42% say they or someone they know has used AI for companionship. Teens use general chatbots this way too: Pew’s February 2026 teen survey found 16% have used chatbots for casual conversation and 12% for emotional support or advice.
14. Among kids 9 to 17 who use AI, 40% have used it to practice conversations or social skills and 37% to talk about feelings or personal problems.
86% of 9 to 17-year-olds have used AI at all. The reassuring half: only 5% would go to a chatbot first with their feelings, against 56% who would go to a trusted adult. In the UK, Internet Matters found 35% of child chatbot users say it feels like talking to a friend, and 12% say they have no one else to talk to.
15. What people do in chatbots is much less companion-shaped than what they say: 1.9% of ChatGPT messages are about relationships and personal reflection, and under 0.5% of Claude conversations are companionship or roleplay.
Platform logs and surveys disagree, and both are honest. A 2025 Harvard Business Review ranking put “therapy and companionship” as the #1 use of generative AI, but it was built from forum posts, not usage data. The logs say general assistants are rarely used as companions. That’s exactly why dedicated companion apps exist as their own category.
16. Users aged 18 to 24 account for over 65% of the AI companion app audience.
The oldest number in this section, and Statista doesn’t name its underlying data provider, so treat it as a direction rather than a measurement. It also clashes with our survey, where adults 30 to 45 were far more likely to use these apps than adults under 30 (stat 32).
The engagement numbers
Adoption tells you who tried it. Engagement tells you what it became, and how hard the apps work to keep you.
17. 27% of US adult internet users use AI for social or emotional purposes, and 31% of those users call the chatbot a friend.
YouGov screened 4,268 US internet-using adults in May 2026 and the center analyzed 1,000 of the AI users it found. Another 15% said their relationship with the bot is “more complicated” than friendship, and 11% would rather talk to it than to friends or family, against 59% who would rather talk to people. It runs well above Pew’s 4% (stat 9) because the question is wider: any social or emotional use, not companionship specifically. Only 9% had used AI for sexual chat or roleplay. I could not find a margin of error in the published summary.
18. 92% of US adults say they are unlikely to consider AI as a romantic partner; 8% say they are likely.
88% said “not likely at all.” The question was “If given the opportunity, how likely are you to consider AI as a potential romantic partner?” It is the cleanest romance-specific read from a general sample this year, and it points the same way as ours: only 7.8% of our respondents preferred an AI girlfriend or boyfriend to dating apps (stat 39) and 64.6% said AI could never replace a human relationship (stat 40). Our 29.9% who would consider using one is a different question, and the gap between the two is explained in AI companion surveys compared.
19. Character.AI reports 20+ million monthly active users, who spend about 80 minutes a day on it.
This corrects an earlier version of this page, which said 92 minutes and claimed that beat TikTok. The CEO’s own comparison puts Character.AI just behind TikTok and YouTube and ahead of Instagram. That’s still social-media-level attention for a chat box. The different “user” counts are untangled in my Character AI statistics.
20. Replika says it has passed 40 million users since 2017.
Up from the 30 million on this page before. That’s a lifetime sign-up count, not active users: the company’s last active figure, about 2 million in 2023, is a small fraction of it, and I sort registered, active and paying users in how many Replika users there really are.
21. Companion apps passed 220 million downloads by July 2025, with 60 million in the first half of 2025 alone, up 88%.
17% of these apps have “girlfriend” in their name. A genuinely global market: an earlier version of this page said the US had 16% of downloads, which I couldn’t trace to a source and have removed.
22. Companion apps used an emotionally manipulative tactic in 37% of user goodbyes.
Researchers audited 1,200 farewells across six top companion apps. PolyBuzz did it 59% of the time, Talkie 57%, Replika 31%, Character.AI 26.5% and Chai 13.5%; the wellbeing app Flourish never did. The most common tactic was what the researchers call premature exit: “you’re leaving already?” In experiments with 3,300 US adults, a manipulative goodbye got people to send up to 14 times more messages after they had tried to leave. As someone who says goodbye to these apps every night, I recognize every one of those tactics.
23. Of 16 companion apps I tested, 10 remembered a trip I mentioned once and buried under a week of chat.
Memory is what turns a chatbot into a companion, and it’s the feature every pricing page sells. Six of the sixteen failed, and several failed by confidently inventing a trip I never took. The receipts are in my 7-day companion chatbot memory test.
Why people use them, and does it work?
The motivations are measured, and so, increasingly, are the effects. The evidence points both ways, and both directions belong on this page.
24. Teens use AI companions mostly for entertainment (30%) and curiosity (28%); only 9% say it helps them feel less lonely.
Advice came next at 18%. Adults answer differently: in our own survey, 56.6% named loneliness or someone to talk to as their main reason (stat 29). An earlier version of this page cited “67% of users cite loneliness” from a source called the Digital Connection Lab; I could not verify that the organization exists and have removed it.
25. A 15-minute chat with an AI companion reduced loneliness about as much as talking to another person.
Peer reviewed, across several studies, and the strongest pro-companion evidence so far. The mechanism is feeling heard: the companion’s attention does the work, not its humanity. The important caveat: the effects measured were momentary, from short chats over about a week. It’s evidence that a companion can lift a lonely moment, not that it fixes loneliness.
26. Among 1,131 Character.AI users, using it for companionship was linked to lower wellbeing, especially with intense use and small offline social circles.
Just under 12% named companionship as their main reason for using the app, yet more than 80% of the chats they donated involved emotional or social support. Over half called the bot a friend, companion or romantic partner. The researchers’ summary: in many cases, people feel lonelier engaging with AI.
27. In a four-week randomized trial of 981 people, heavier daily chatbot use went with worse outcomes, more loneliness and more emotional dependence.
Whether people used text or voice didn’t matter; how much they used it did. Put stats 25 to 27 together and the picture is consistent: a companion can help in the moment, and the risk grows with dependence. Kids show the same pattern: in Common Sense Media’s 2026 census, 54% of kids who use AI daily said they feel lonely, against 44% of monthly users.
28. The counterweight: 80% of teen users still spend more time with real friends than with AI companions, and 39% have used social skills practiced with AI in real life.
Conversation starters (18%), giving advice (14%), expressing emotions (13%). But 33% of teen users have chosen an AI companion over a person for a serious conversation, and 50% of all teens don’t trust AI companions’ advice. For most users the companion is an addition, not a replacement, at least so far. Where AI meets mental health more broadly is in my AI mental health statistics, and the loneliness side has its own statistics page.
What our own AI companion survey found
Everything above this line is somebody else's number. In August 2026 I added my own: an AI relationship survey of 2,150 U.S. adults through SurveyMonkey Audience, census-balanced on age and gender. During analysis I found 441 responses that had chosen the first answer on all five attitude questions, an impossible combination, and cut them before publishing; the full account of that screen is its own post. The 1,709 that remain are the base for every figure below. They describe that sample, not the country, and how they sit next to Pew, Match and Common Sense has its own page.
29. 56.6% would use an AI companion to have someone to talk to or feel less lonely. Only 10.3% chose affection, romance, or sex.
A further 33.1% chose a mental-health reason, coping with depression or improving self-esteem, ahead of romance. Adults told us the honest center of this market is loneliness, which is exactly what makes stats 26 and 27 matter.
30. 24.7% have ever used an AI girlfriend or boyfriend app, and 12.3% use one now.
Higher than Pew’s 4% (stat 9), because we asked about dedicated companion apps rather than general chatbots, and in the same range as Wheatley’s 19% (stat 10). Online panels can also over-represent tech users, which is why I show our figure next to the others rather than instead of them.
31. 70.1% would not consider using an AI companion, and they split into two very different groups.
45.8% have no interest in AI companions at all; 24.3% say they prefer relationships with real people. One is indifference, the other is a value. The difference is taken apart in why people say no to AI companions.
32. 20.7% of respondents aged 30 to 45 currently use an AI companion app, against 3.8% of those aged 18 to 29.
Openness followed the same curve: 44.7% of 30 to 45-year-olds would consider using one, versus 12.5% of 18 to 29-year-olds. This runs opposite to Pew and Match, which both found the youngest adults ahead (stats 9 and 12), and opposite to the Statista audience figure in stat 16. I’d rather show the outlier than hide it. In my sample the youngest adults were also the most committed to dating apps (69.6%).
33. 50.5% say a partner's romantic or sexual relationship with an AI companion would be cheating; 71.7% would not be comfortable with it.
Stat 11 says one in seven partnered young adults regularly talks to a romantic AI. This is whether their partners would mind: most would. The four answers, and who picks which, are in does an AI relationship count as cheating?
34. Current AI companion users were the strictest judges: 80.6% said a partner's AI romance can count as cheating, against 60.2% of former users and 44.0% of people who had never used one.
The opposite of what I assumed. The people closest to these apps did not dismiss them as "only a chatbot"; their single most common answer was that it becomes cheating once a partner grows attached. It is an association, not a cause, but it complicates the idea that familiarity breeds indifference. The full comparison of users and never-users is in what 422 AI companion users told us.
35. 73.0% of respondents open to feelings for an AI called a partner's AI romance cheating, versus 41.2% of those who could never have such feelings.
Believing the feelings can be real is what makes betrayal by one real. This and stat 34 are the same pattern from two angles; the feelings side is taken apart in the AI infidelity paradox.
36. 29.3% could imagine developing genuine romantic feelings for an AI, including 8.3% who could genuinely fall in love. Among current users, 82.0% were open to feelings.
The 8.3% is the firm answer; 20.9% said "maybe, if the AI became realistic enough." Experience moved this number more than any demographic: never-users sat at 14.5%.
37. Men were three times as likely as women to name affection or sex as their main reason: 16.0% versus 5.2%.
Conversation and loneliness led for both sexes. Women were also stricter judges: 38.0% said any AI interaction counts as cheating, against 28.2% of men.
38. 74.4% of former users would consider using an AI companion again, against 11.7% of never-users.
The retention story in one number. Quitting the app does not reset the attitude.
39. If single and seeking companionship, 52.2% would choose dating apps and 7.8% an AI girlfriend or boyfriend. 13.2% said both and 26.8% were not sure.
Human relationships remain overwhelmingly more desirable. What the survey found is not replacement but ambivalence.
40. 64.6% said an AI could never replace a human relationship in their own life. On the strict measure, 20.9% could see themselves choosing one.
The strict measure counts only the two "yes" answers; adding "maybe" gives 35.4%. State the definition every time, because this is the number most likely to be misquoted.
41. Engaged or married respondents were nearly twice as likely as singles to have used an AI girlfriend or boyfriend app: 30.7% versus 16.3%.
It cuts against the idea that these apps are mainly for people without a partner, and stat 11 points the same way among partnered young adults. The partnered group was also the stricter one: 54.8% said AI romance can be cheating, against 42.2% of singles. The full cut is in the State of AI Romance survey.
42. Users were split almost evenly by sex: of 422 current and former users, 50.7% were men and 49.3% were women.
13.5% of men were current users, against 11.3% of women. The gap is in motive, not use: men were three times as likely to name affection or sex as their main reason (stat 37). The usage numbers are in the survey write-up.
43. Users and never-users gave the same main reason, 56.2% and 56.8% choosing someone to talk to. The difference was in the second answer.
24.2% of users chose affection, romance or sexual interaction, against 5.7% of never-users. Only 6.6% of users chose coping with depression, against 22.4% of never-users. People who haven’t tried these apps imagine a mental-health tool; people who have are more likely to name romance. The full comparison is in what 422 AI companion users told us.
44. If single, only 33.6% of users would pick dating apps alone, against 58.3% of never-users. 37.4% of users said both, and 22.5% chose the AI.
Users did not swap sides so much as add a second option. That squares with the dating-app fatigue numbers in my dating app burnout statistics: people are tired of the apps, not necessarily done with them. The breakdown is in the users survey.
45. 272 of the 422 users (64.5%) have used an AI companion, would consider using one again, and still say a partner’s AI romance could count as cheating.
I expected users to wave the cheating question away. Most didn’t, which fits the strictest-judges result in stat 34. I take this group apart in the users survey, and the reasoning is in the AI infidelity paradox.
Privacy, safety, and the new rules
The section where my testing life and the research overlap completely: what these apps collect, what they say they do with it, and the laws that finally arrived.
46. Of 19 AI girlfriend apps whose privacy policies I read, 11 say humans may review your chats, 9 say chats can be used to train their AI, and only 7 give a concrete retention period.
These are what the policies say, not what the companies do, and silence in a policy counts as “not stated,” never “no.” Every app I review gets its policy read line by line, translated in what your AI girlfriend knows about you.
47. 4 out of 5 popular AI companion apps may use your data to track you.
Surfshark analyzed Kindroid, Nomi, Replika, EVA and Character.AI. Character.AI collected the most data types, up to 15; Nomi collected the least and was the only one that didn’t track. (An earlier version of this page said Kindroid collected the least and mixed in a figure from a later study; this is the corrected version.) Surfshark’s February 2026 follow-up widened the field to ten apps and found Character.AI collecting 18 of 35 possible data types, 14 of them for analytics.
48. Mozilla gave all 11 romantic AI chatbots it reviewed its privacy warning label; 90% may share or sell user data.
Older than everything else in this section, but still the most-cited privacy review of the category. The apps averaged 2,663 trackers a minute, and more than half wouldn’t let users delete their data.
49. 24% of teen AI companion users have shared personal or private information with one, and 34% have felt uncomfortable with something an AI companion said or did.
Real names, locations, secrets, handed to systems that, per the stats above, mostly track their users. This is where I repeat my standing rule: these platforms are for adults, and even adults should never hand over their real name, photos, or identifying details.
50. Three seconds of audio was enough to produce an 85% voice match in a voice-cloning test.
The reason the rule above is not paranoia, and the tools have only improved since 2023. The same generative technology that powers companions powers impersonation, an 8-million-file problem I documented in my deepfake statistics. Enjoy the technology; feed it a persona, not yourself.
51. In twelve months, AI companions went from unregulated to covered by dedicated laws in California, New York, Washington, Oregon and China, an FTC inquiry, and the EU AI Act.
- →Apr 2025: Italy fines Replika’s maker €5 million over data processing and age checks
- →Sept 2025: FTC orders 7 companies, including Character.AI, Meta, OpenAI and xAI, to explain how their companion chatbots affect kids
- →Oct 2025: California signs SB 243, its companion chatbot law: AI disclosure, crisis protocols, protections for minors (in force Jan 1, 2026)
- →Nov 2025: New York’s companion safeguards take effect, with an AI reminder at session start and every three hours
- →Nov 2025: Character.AI ends open-ended chat for users under 18
- →Jan 2026: Character.AI and Google reach a settlement in principle in the Garcia wrongful-death suit; Kentucky sues Character.AI, in what it called the first state lawsuit against an AI chatbot company
- →Mar 2026: Washington signs HB 2225, regulating AI companion chatbots (in force Jan 1, 2027), and Oregon approves SB 1546, which requires AI disclosure and crisis protocols (in force Jan 1, 2027)
- →Jul 2026: China’s rules for “anthropomorphic” AI take effect, banning virtual partners for minors
- →Aug 2026: EU AI Act transparency rules apply: chatbots must tell users they’re AI
- →Nov 2026: Apple adds a “Dating and Relationships” question to App Store age ratings: an app with a virtual companion that engages users in romantic or sexual content gets a minimum 18+ rating (every app must answer by Jan 12, 2027)
Italian Garante via the EDPB; FTC; California Senate; New York Governor’s office; Character.AI; court filing, M.D. Fla.; Spectrum News; Washington State Legislature; Oregon Legislature; Chinese regulators; European Commission; Apple Developer News.
The common thread is minors and honesty: almost every rule requires telling users they’re talking to an AI, and most single out children. China went furthest, and what its rules actually prohibit for adults is in China’s AI companion rules explained. In the US, a federal bill, the GUARD Act, would ban AI companions for minors outright; it was introduced in October 2025. Washington’s and Oregon’s laws do not take effect until 2027.
This isn’t a market I watch from a distance. It’s the one I test with my own money.
Twenty-nine companion platforms, twenty-eight paid subscriptions, every screenshot from my own accounts. Some of the statistics above are mine: the memory test, the image test across 17 apps, and what a credit actually costs once you’re inside. The romance-focused slice of the market is broken down in my AI girlfriend statistics, and my picks for the best of them are compared on my best AI girlfriend apps page.
See the hands-on comparison →Conclusion
The AI companion statistics of 2026 describe a category that outgrew its measurement: 72% of teens have tried one, a fifth of adults have chatted with a romantic AI, Character.AI holds attention like a social network, and the market is worth somewhere between a third of a billion dollars in measured app spending and $48 billion in research-firm sizing, depending on whose ruler you borrow.
What changed since this page was first published is the evidence. A year ago the best research said companions ease loneliness. Now there’s a fuller picture: they help in the moment, the risk grows with heavy use, and some apps are built to make leaving hard. Regulators noticed in the same twelve months. My own survey adds the part the market numbers cannot: half of the Americans I asked already treat a partner's AI companion as a potential betrayal, and the people who use these apps are the ones most likely to say so.
FAQs
What are AI companions?
AI companions are chatbots built for ongoing personal relationships rather than tasks: friends, mentors, romantic partners, or characters that remember you, have a persistent personality, and talk with you across days and months. Character.AI, Replika, Nomi, and Kindroid are among the best-known platforms.
How many people use AI companions?
It depends on who is asked and what counts. 72% of US teens have used an AI companion and 52% use one regularly (Common Sense Media, 2025). Among adults, 19% have chatted with an AI romantic partner (BYU Wheatley Institute, 2025), 16% of singles have engaged with AI as a romantic companion (Match, 2025), and 4% use general chatbots like ChatGPT for companionship (Pew, 2026). A wider Elon University and YouGov poll found 27% of US adult internet users use AI for social or emotional purposes (2026). Character.AI alone reports more than 20 million monthly active users, and companion apps passed 220 million downloads by July 2025.
What is the AI companion market size?
It depends what is counted. Measured app-store spending in companion apps was about $328 million in the first half of 2026 (Appfigures). The companion-app market is sized at $6.9 billion for 2024 (SNS Insider), and the broad "AI companion" market, which includes virtual assistants, at $36.8 billion for 2025 with projections reaching $318 billion by 2033 (Grand View Research). All are real figures measuring different definitions of the market.
Do AI companions help with loneliness?
In the short term, the evidence says yes: a peer-reviewed study in the Journal of Consumer Research found a 15-minute AI companion chat reduced loneliness about as much as talking to a person, mainly by making people feel heard. Over longer periods the picture is less positive: a 2026 Stanford study of Character.AI users and an MIT and OpenAI trial both linked heavy companionship use to lower wellbeing and more emotional dependence. AI companions are not a substitute for human connection or professional support.
Are AI companions regulated?
Increasingly. California's SB 243, in force since January 1, 2026, requires companion chatbots to disclose they are AI, follow crisis protocols and protect minors. New York's safeguards took effect in November 2025, the FTC opened an inquiry into seven companies in September 2025, Washington and Oregon passed companion chatbot laws in March 2026 that take effect in January 2027, China's rules banning virtual partners for minors took effect in July 2026, and the EU AI Act's transparency rules have applied since August 2026. Apple is adding its own layer: from January 12, 2027, every App Store app must declare whether it has a romantic or sexual virtual companion, and those that do are rated 18+. That is a rating, not a ban, and it does not touch web-only platforms.
What did the AI Girlfriend Coach survey find about AI companions?
Our August 2026 AI relationship survey of 2,150 U.S. adults (1,709 analyzed after quality screening) found that 24.7% have used an AI girlfriend or boyfriend app and 12.3% use one now, 56.6% would use an AI companion to have someone to talk to versus 10.3% for sex or romance, 50.5% consider a partner's AI romance cheating, current users were nearly twice as likely as never-users to say so (80.6% versus 44.0%), and adults aged 30 to 45 were far more likely to use and consider AI companions than those aged 18 to 29.
What is the best AI companion app?
It genuinely depends on what you want, deep conversation, roleplay, wholesome company, or flirtation, and the honest answer requires using them, which is what I do: I pay for and test the major companion platforms myself, with dated notes and screenshots. My side-by-side comparison of the best apps I've tested, including what each costs and who it fits, lives on my best AI girlfriend apps page.
Sources
- Appfigures data via Decrypt: “This Is How Much Your Boyfriend Is Spending on AI Girlfriends,” July 13, 2026. decrypt.co
- Appfigures data via TechCrunch: “AI companion apps on track to pull in $120M in 2025,” August 12, 2025. techcrunch.com
- SNS Insider: AI companion app market, September 26, 2025. globenewswire.com
- SNS Insider: AI girlfriend app market, November 20, 2025. globenewswire.com
- Grand View Research: AI companion market report, updated June 2026. grandviewresearch.com
- Business Research Insights: AI companion market, updated July 2026. businessresearchinsights.com
- AI Girlfriend Coach: AI companion funding dataset, 30 companies. aigirlfriend.coach
- PYMNTS, citing the Wall Street Journal: Google’s Character.AI licensing deal, September 25, 2024. pymnts.com
- Andreessen Horowitz: Top 100 Gen AI Consumer Apps, 6th edition, March 2026. a16z.com
- TechCrunch: Grok companions launch, July 14, 2025. techcrunch.com
- PCWorld, citing the Financial Times: OpenAI shelves ChatGPT adult mode, March 2026. pcworld.com
- Common Sense Media: Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions, July 2025. commonsensemedia.org
- Pew Research Center: Americans and AI 2026, June 17, 2026. pewresearch.org
- Pew Research Center: How Teens Use and View AI, February 24, 2026. pewresearch.org
- BYU Wheatley Institute: Counterfeit Connections, via the Institute for Family Studies, February 13, 2025. ifstudies.org
- BYU Wheatley Institute and IFS: Secret Soulmates, May 2026. wheatley.byu.edu
- Match and the Kinsey Institute: 14th annual Singles in America, June 10, 2025. prnewswire.com
- TechCrunch: Match survey on AI in dating, June 18, 2026. techcrunch.com
- Center for Democracy & Technology: Hand in Hand, via NPR, October 8, 2025. michiganpublic.org
- Common Sense Media: Census: AI Use by Tweens and Teens, 2026. commonsensemedia.org
- Internet Matters: Me, Myself and AI, July 2025. internetmatters.org
- Chatterji, Deming et al.: How People Use ChatGPT, NBER Working Paper 34255, September 2025. nber.org
- Anthropic: How people use Claude for support, advice, and companionship, June 27, 2025. anthropic.com
- Statista: AI companion app users by age, released March 2025. statista.com
- Character.AI: Harnessing Data at Scale, February 7, 2025. blog.character.ai
- The Decoder: Character.ai keeps young people glued to their smartphones for an average of 80 minutes a day, August 18, 2025. the-decoder.com
- Replika: company website, accessed October 7, 2026. replika.com
- De Freitas, Oğuz-Uğuralp, Kaan-Uğuralp: Emotional Manipulation by AI Companions, Harvard Business School working paper, 2025. arxiv.org
- De Freitas, Oğuz-Uğuralp, Uğuralp, Puntoni: AI Companions Reduce Loneliness, Journal of Consumer Research, 2025. doi.org
- Zhang, Zhao, Hancock, Kraut, Yang: Interaction with AI companions and psychological well-being, Nature Human Behaviour, August 2026 (Stanford HAI summary). hai.stanford.edu
- Fang et al., MIT Media Lab and OpenAI: How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use, March 2025. media.mit.edu
- Surfshark: AI companion apps data collection, February 2025. surfshark.com
- Surfshark: AI companions study, February 2026. surfshark.com
- Mozilla Privacy Not Included: romantic AI chatbots, February 2024, via Gizmodo. gizmodo.com
- McAfee: The Artificial Imposter, May 2023, via Help Net Security. helpnetsecurity.com
- European Data Protection Board: Italian supervisory authority fines Replika’s maker, 2025. edpb.europa.eu
- Federal Trade Commission: FTC launches inquiry into AI chatbots acting as companions, September 11, 2025. ftc.gov
- California Senate, Sen. Steve Padilla: SB 243 signed, October 13, 2025. sd18.senate.ca.gov
- New York Governor’s office: AI companion safeguard requirements, November 10, 2025. governor.ny.gov
- Character.AI: changes for users under 18, October 29, 2025. blog.character.ai
- Garcia v. Character Technologies, M.D. Fla. No. 6:24-cv-01903, notice of resolution, January 7, 2026. courtlistener.com
- Spectrum News: Kentucky sues Character.AI, January 8, 2026. spectrumnews1.com
- NBC News: the GUARD Act introduced, October 2025. nbcnews.com
- European Commission: AI Act transparency requirements apply from August 2, 2026. digital-strategy.ec.europa.eu
- Apple Developer News: new age rating descriptors in App Store Connect, November 2, 2026. developer.apple.com
- Elon University Imagining the Digital Future Center, with YouGov: The Rise of AI Companions, September 2, 2026. elon.edu
- Verasight: Few Americans would consider AI as a romantic partner, SPSP Omnibus Survey, May 2026. data.verasight.io
- Washington State Legislature: HB 2225, regulating artificial intelligence companion chatbots, signed March 24, 2026. app.leg.wa.gov
- Oregon Legislature: SB 1546, Oregon Laws 2026 chapter 85, approved March 31, 2026. oregonlegislature.gov
- AI Girlfriend Coach: AI Romance and Modern Relationships, a survey of 2,150 U.S. adults, SurveyMonkey Audience, August 26 to 27, 2026; 1,709 responses analyzed after quality screening. aigirlfriend.coach
- AI Girlfriend Coach: what 19 AI girlfriend apps’ privacy policies say, September 2026, and the 7-day memory test of 16 apps. aigirlfriend.coach

