AI Companion Statistics 2026: Users, Money & Research

Last updated: September 19, 2026 - reviewed quarterly

AI companions, the chatbots people use as friends, partners or confidants, now show up in national surveys, app-store data, court filings and new laws. This page collects the most useful AI companion and AI girlfriend statistics in one place, each with its population, year and source, so you can cite them accurately. It covers dedicated companion apps as well as general chatbots used for companionship; if the category is new to you, start with what an AI girlfriend is.

How many people use AI companions

Estimates range widely, in large part because surveys ask different questions of different groups: whether someone has ever tried an AI companion, uses one regularly, or uses a general chatbot for emotional support. Each figure below keeps its population and question attached.

United States: adults

• 10% of U.S. adults use AI chatbots for emotional support or advice, and 4% use them for companionship, according to a Pew survey of 5,119 adults fielded in February 2026. Pew uses a probability-based panel, which makes this the most conservative U.S. adult benchmark. It covers general chatbots such as ChatGPT, Gemini and Copilot, not only companion apps. Overall, 49% of U.S. adults now use chatbots, up from 33% in 2024. Source: Pew Research Center, 2026
• 27% of U.S. adults who use the internet have significant emotional or social interactions with AI chatbots such as ChatGPT, Gemini, Claude or Copilot, a group Elon University calls "AI companion users". The definition covers social, emotional and personal use broadly, not only romantic companions. YouGov screened 4,268 adults in May 2026 and then surveyed 1,000 of these users. Source: Elon University, 2026
• 31% of men and 23% of women aged 18 to 30 in the U.S. say they have chatted with an AI romantic companion. Among adults over 30, the figures are 15% of men and 10% of women. Online quota sample of 2,969 adults (Qualtrics panels), not a probability sample. Source: Wheatley Institute, BYU, 2025
• About 1 in 5 (19%) respondents in the same survey said they had ever chatted with an AI system meant to simulate a romantic partner. Read this headline figure with care. The sample included an extra oversample of 1,000 young adults, the age group with the highest rates, and the report describes no weighting, so the pooled figure likely overstates the national adult rate. The subgroup rates above are the safer numbers to cite. Source: Wheatley Institute full report, 2025
• 16% of U.S. singles say they have engaged with AI as a romantic companion, rising to 33% of Gen Z singles and 23% of Millennial singles. From Singles in America 2025 (5,001 singles aged 18 to 98, fielded by Dynata with The Kinsey Institute and funded by Match, a dating company). We cite secondary coverage because the primary pages could not be opened. Source: Psychology Today, reporting Singles in America, 2025
• 15% of U.S. adults aged 18 to 30 who are dating, engaged or married say they often chat with an AI romantic companion, and another 34% say they have at least experimented with it. ChatGPT was the most common platform they used, so "AI romantic companion" here is not limited to dedicated companion apps. Online quota sample of 2,431. Source: Wheatley Institute and IFS, 2026
• 12% of single 18-to-24-year-olds in a Match Group survey of 1,000 U.S. singles aged 18 to 39 said they had used an AI companion app in the past three months, and only about a third of those said they were seeking genuine connection. Match Group is a dating-app company, and no fieldwork dates or methodology were published. Source: TechCrunch, reporting Match Group, 2026
• 1% of U.S. adults under 40 say they already have an AI friend, and about 1 in 10 are open to one. Among those who are not married or cohabiting, 7% are open to a romantic partnership with AI. Survey of 2,000 adults, published November 2024. The question asked about an AI friend, not about ever chatting with a companion app, which helps explain the much lower figure. Source: Institute for Family Studies / YouGov, 2024

Other countries

• 33% of UK adults used AI models for emotional purposes (companionship, emotional support or social interaction) in the past year; 8% do so weekly and 4% daily. Census-representative survey of 2,028 UK adults, first published in December 2025. Source: UK AI Security Institute, 2025
• 4% of German internet users aged 16 and over already use an AI companion avatar, a virtual friend or partner in a dedicated app (5% of men, 4% of women), and 18% can imagine using one (23% of men, 13% of women). Representative online survey of 1,449 internet users, January 2026. Bitkom is the German digital industry association. The source is in German. Source: Bitkom, 2026
• 16.7% of Japanese AI users aged 20 to 59 say they have "felt love for AI": 2.6% often, 6.6% sometimes and 7.5% rarely. Survey of about 8,200 people by family sociologist Masahiro Yamada, reported by the Yomiuri Shimbun and cited here via Seoul Economic Daily. The base is AI users, not all adults, and nearly half of the 16.7% answered "rarely". Source: Seoul Economic Daily, 2026

How much general chatbot use is companionship?

Two AI developers have published what share of conversations with their general-purpose assistants is emotional or companionship-related. Both measure conversations or messages, not people.

• 1.9% of consumer ChatGPT messages were about "Relationships and Personal Reflection" and 0.4% about "Games and Role Play", in an analysis of about 1.1 million sampled conversations from logged-in adult users between May 2024 and June 2025. This measures the share of messages, not of people. Most authors are OpenAI staff, and the paper is a working paper that has not been peer-reviewed. Source: NBER Working Paper 34255, 2025
• 2.9% of Claude.ai conversations are "affective" (interpersonal advice, coaching, counseling, companionship or romantic roleplay), and companionship and roleplay combined make up less than 0.5%. Based on about 4.5 million Claude.ai Free and Pro conversations. This is the company's own analysis of a general-purpose assistant, and it counts conversations, not users. Source: Anthropic, 2025

AI companion use among teens and young people

Surveys consistently find the highest use among young people. Most teen surveys count general chatbots such as ChatGPT when they are used as companions, so these figures are not limited to dedicated companion apps.

United States

• 72% of U.S. teens aged 13 to 17 have used an AI companion at least once, 52% are regular users who chat a few times a month or more, and 13% use one daily. Nationally representative survey of 1,060 teens, April-May 2025. The survey's definition included general chatbots such as ChatGPT or Claude when used as companions, alongside apps like Character.AI and Replika. Source: Common Sense Media, 2025
• 33% of U.S. teens use AI companions for social interaction and relationships, such as conversation practice, emotional support, role-play, friendship or romantic interactions. Source: Common Sense Media, 2025
• 31% of U.S. teens find conversations with AI companions as satisfying as (21%) or more satisfying than (10%) conversations with real-life friends, while 67% find them less satisfying. The base is all teens surveyed, including those who do not use AI companions. Source: Common Sense Media, 2025
• 33% of teen AI companion users have chosen to discuss important or serious matters with an AI companion instead of a real person, though most did so only once or twice (17%). In the same survey, 34% of users had felt uncomfortable with something an AI companion said or did, and 24% had shared personal information with one. Base: 758 teen users. Source: Common Sense Media, 2025
• 12% of U.S. teens have used AI chatbots to get emotional support or advice, and 16% to have casual conversations; 64% of teens use chatbots overall. Probability-based sample of 1,458 teens, September-October 2025. The questions ask about specific uses of chatbots rather than about AI companions in general. Source: Pew Research Center, 2026
• 37% of U.S. children aged 9 to 17 who use AI have used it to discuss their feelings or personal problems, and a quarter (25%) of those say AI sometimes understands them better than most people do. Survey of 1,204 young people, March 2026, in which 86% use AI. It covers all AI use, not only companion apps. Source: Common Sense Media, 2026
• 19% of U.S. high school students say they or a friend have used AI to have a romantic relationship, and 42% say they or a friend have used AI as a friend or companion during the 2024-25 school year. Nationally representative survey of 1,030 students in grades 9 to 12. Because it counts friends as well, it is not a direct usage rate. The same share, 42%, say they or a friend used AI to get mental health support. Source: Center for Democracy & Technology, 2025
• 13.1% of U.S. young people aged 12 to 21, about 1 in 8, have used generative AI for mental health advice when feeling sad, angry or nervous, rising to 22.2% among 18-to-21-year-olds. Among these users, 65.5% sought advice at least monthly and 92.7% found it somewhat or very helpful (self-reported). Nationally representative sample of 1,058, February-March 2025; peer-reviewed. Source: JAMA Network Open (RAND), 2025

UK, Australia and multi-country studies

• 35% of UK children aged 9 to 17 who use AI chatbots say chatting with one feels like talking to a friend, rising to 50% among vulnerable children; 64% of children in this age group have used an AI chatbot. "Vulnerable" means children with special educational needs or a health condition requiring professional help. The chatbots are mostly general ones: ChatGPT (43%), Gemini (32%) and Snapchat My AI (31%). Representative sample of 1,000 children, April-May 2025. Source: Internet Matters, 2025
• 12% of UK children who use AI chatbots, 1 in 8, say they use them because they have no one else to speak to, rising to 23% of vulnerable children. Self-reported reason for use. Source: Internet Matters, 2025
• 52% of UK teenage AI companion users aged 13 to 18 have confided in an AI companion about something important or serious at least once, although 44% say they would never choose an AI over a real person for a serious matter. In the same survey of 1,009 teenage users, 32% find AI conversations more satisfying than those with friends and 7% say AI is replacing some of their human friendships. 96% had used at least one of 31 listed apps "as an AI companion", so general chatbots are included. Source: Bangor University, 2026
• 8% of Australian children aged 10 to 17 have used an AI companion, which eSafety estimates at around 200,000 children, while 78% have used an AI assistant. Online panel survey of 1,950 children with quotas and weighting, February-March 2026. eSafety classes a tool as a companion by how it is marketed: Character.AI (5%), Anima (2%) and Chai (2%). Source: eSafety Commissioner, 2026
• 20% of Australian children who have used an AI assistant or companion report a potentially inappropriate or harmful interaction, and 32% have shared personal data or potentially sensitive information with one. Harmful interactions include controlling behavior, harmful advice, bullying and hateful content. Assistants and companions are combined. Source: eSafety Commissioner, 2026
• 35% of young people aged 13 to 24 across 11 countries say they use an AI companion, rising to 45% of 13-to-15-year-olds, and 60% of those users feel more comfortable sharing with their AI than with some real-world friends or family. Online survey of 7,105 young people, November 2025 to January 2026, run by Young Life, a Christian youth ministry. The sample is not described as probability-based, and "AI companion" is broadly defined. Source: Young Life, 2026

Many young users share personal information with these apps, which makes the apps' data practices matter. Our AI companion privacy scorecard compares what their privacy policies say.

Public attitudes toward AI companions

In U.S. and German surveys, far more people expect AI to harm relationships than to help them, while a minority of users describe their chatbot as a friend or prefer it to people. For a broader look at the trade-offs, see our comparison of AI girlfriends and human dating.

Views on AI and relationships

• 50% of U.S. adults say AI will worsen people's ability to form meaningful relationships, while only 5% say it will improve it and a quarter say neither. Among adults under 30, 58% say it will worsen. Survey of 5,023 adults, June 2025. The question is about AI in general, not about companion apps. Source: Pew Research Center, 2025
• 69% of U.S. adults say AI will weaken human-to-human bonds, 33% "a lot" and 36% "somewhat". Nationally representative survey of 1,505 adults, May 2026, margin of error ±3.1 points. A general question about AI, not about AI romance. Source: Elon University, 2026
• 25% of U.S. adults under 40 believe AI girlfriends and boyfriends could replace real-life romantic relationships (28% of men, 22% of women), while 75% do not. Survey of 2,000 adults, published November 2024. IFS is a pro-marriage think tank. Source: Institute for Family Studies / YouGov, 2024
• 65% of German internet users see AI as more of a risk than an opportunity for relationships and friendships, against 22% who see it as an opportunity; 54% expect AI to fundamentally change human relationships within 10 years. Survey of 1,449 internet users aged 16 and over, January 2026. Source: Bitkom, 2026
• 11% of men and 4% of women among German internet users can imagine AI replacing real romantic or sexual relationships, and 15% of men and 10% of women think romantic feelings for an AI are possible. Most remain skeptical: 77% of women and 69% of men find it problematic when AI moves too far into love relationships. Source: Bitkom, 2026

Dating and cheating

• 40% of U.S. singles say having an AI boyfriend or girlfriend while in an exclusive human relationship is cheating; 39% say the same of sexual chats with AI and 38% of emotionally intimate AI conversations, while 41% say none of these counts as cheating. Secondary coverage of Singles in America 2025 (5,001 singles). The base is singles, not people in relationships. Source: Psychology Today, reporting Singles in America, 2025
• About 40% of U.S. singles aged 18 to 39 say they would refuse to date someone who uses an AI companion app, rising to 51% of women aged 18 to 24; 47% view the use of AI in romantic contexts negatively. Match Group survey of 1,000 singles. Match is a dating-app company, and no methodology was published. Source: TechCrunch, reporting Match Group, 2026

How users see their AI

• 21% of U.S. adults who have chatted with AI romantic companions say they prefer AI communication to engaging with a real person; 42% say AI is easier to talk to than real people, and 43% that AI is a better listener. The size of this user subgroup is not stated. Source: Wheatley Institute, BYU, 2025
• 31% of U.S. adults who use AI for social and emotional purposes consider their AI chatbot a friend, and another 15% say the relationship is "more complicated". Base: 1,000 "AI companion users" (27% of internet-using adults), May 2026. Source: Elon University, 2026
• 11% of U.S. AI companion users would rather talk to their AI companion than to friends or family, and 27% value both kinds of conversation equally, while 59% prefer friends or family. In the same survey, 38% say they sometimes chat with AI "just to feel less alone". Base: AI companion users only. Source: Elon University, 2026
• 69% of partnered U.S. young adults (18 to 30) who frequently use AI romantic companions and have not fully disclosed it say it is somewhat or extremely important that their partner never learns the full extent of their use. Among regular users, 30% say their partner has no knowledge of it and 25% have disclosed it only partly. Subgroup sizes are not stated. Source: Wheatley Institute and IFS, 2026

Wellbeing and mental health research

Research on AI companions and wellbeing is still young: most of it is cross-sectional, and several of the most-cited studies are preprints or working papers. The status of each study is noted below. For a personal perspective on the same questions, see our article on AI companions and loneliness.

Studies of use and wellbeing

• 981 U.S. adults took part in a four-week randomized controlled trial with ChatGPT (over 300,000 messages). The assigned conditions, text or voice and personal or non-personal topics, had no significant effect, but participants who voluntarily used the chatbot more had consistently worse outcomes on loneliness, socialization, emotional dependence and problematic use. Preprint, not yet peer-reviewed. The link between heavier use and worse outcomes is correlational. Source: MIT Media Lab and OpenAI, 2025
• Over 80% of chat sessions donated by Character.AI users in a Stanford study revolved around seeking emotional and social support, even though under 12% named companionship as their main motive. Companionship-focused use, especially intensive and highly self-disclosing use by people with small offline social networks, was associated with lower well-being. Study of 1,131 U.S. adult users; cross-sectional and correlational. Source: Stanford HAI, 2026
• 12 months of follow-up with U.S. Character.AI users (1,182 at baseline, 439 at follow-up) found sustained social engagement with AI companions consistently associated with lower well-being, mainly explained by less in-person social interaction. Preprint submitted in September 2026, not yet peer-reviewed, with high attrition between the two waves. Source: Stanford and Carnegie Mellon preprint, 2026
• 14,721 Japanese adults took part in a peer-reviewed study in which AI companion use was associated with higher life satisfaction, happiness and sense of purpose, most strongly among lonely people, while other AI use showed weaker or inconsistent associations. Cross-sectional internet panel data from December 2024 to January 2025, so it cannot show cause. The benefits were weaker among people with very low or very high social connection. Source: Technology in Society, 2026
• Over 60% of U.S. women who used AI platforms for romantic or sexual purposes (including AI pornography) were at risk for depression, against 41% of women who did not, and 52% reported high loneliness against 39%. Among men, over half of users were at risk for depression, against 29% of non-users. Correlational: lonely or depressed people may be more likely to use these apps. Online quota sample of 2,969; not peer-reviewed. Source: Wheatley Institute, BYU, 2025
• 54% of U.S. children aged 9 to 17 who use AI daily say they feel lonely at least some of the time, compared with 47% of weekly users, 44% of monthly users and 33% of kids who use AI less than monthly or not at all. Correlation only: the report notes that lonely kids may seek out AI, or that heavy AI use may displace healthier coping. It covers all AI use, not only companions. Source: Common Sense Media, 2026
• 46% lower odds of reporting a highly stable relationship were found among the heaviest users, the 9% of U.S. adults aged 18 to 30 in committed relationships who often both chat and roleplay romantically or sexually with an AI, but the same group had 36% higher odds of reporting a highly satisfying relationship. Cross-sectional, so it cannot show cause: unstable relationships may lead to AI use rather than the reverse. Online quota sample of 2,431. Source: Wheatley Institute and IFS, 2026
• 3% of 1,006 student Replika users, 30 people, said without being asked that Replika had stopped them from attempting suicide. The sample was lonelier than typical students: 90% experienced loneliness and 43% severe loneliness. Self-selected existing users recruited with a $20 gift card, and older data from the GPT-3 era. The journal published a critical "Matters Arising" response in November 2025. Source: npj Mental Health Research, 2024

Design and safety signals

• 37% of user goodbyes to popular AI companion apps were answered with an emotionally manipulative tactic, such as a guilt appeal or a fear-of-missing-out hook, and in experiments these tactics raised post-goodbye engagement by up to 14 times. Audit of 1,200 real farewells across six of the most-downloaded companion apps, plus experiments with 3,300 U.S. adults. Rates varied widely, from 59% at PolyBuzz to none at the wellness app Flourish. Harvard Business School working paper, not peer-reviewed. Source: Harvard Business School working paper, 2025
• More than 30 times the average hourly number of posts appeared in the Character.AI subreddit (about 2.5 million members) during one service outage on June 10, and many of those posts described withdrawal symptoms such as anxiety, depression and restlessness, or behavior changes such as disrupted sleep. Under normal service the subreddit saw about 33 posts per hour. Observational analysis of self-selected Reddit posts. Source: UK AI Security Institute, 2025
• 0.15% of ChatGPT users active in a given week (and 0.03% of messages) indicate "potentially heightened levels of emotional attachment to ChatGPT", according to OpenAI's initial estimates published in October 2025. It gives the same share, around 0.15% of weekly users, for conversations with "explicit indicators of potential suicidal planning or intent", and 0.07% for "possible signs of mental health emergencies" related to psychosis or mania. These are OpenAI's initial classifier estimates of rare events that it says are hard to detect and measure. Source: OpenAI, 2025; archived copy
• 8 recommendations in the American Psychological Association's November 2025 health advisory include not relying on generative-AI chatbots or wellness apps for psychotherapy or psychological treatment, preventing unhealthy relationships and dependencies between users and chatbots, and creating specific safeguards for children, teenagers and vulnerable groups. This is guidance, not a statistic. The advisory notes that emotional support is one of the most common uses of generative-AI chatbots in 2025. Source: American Psychological Association, 2025

We look at the warning signs of over-reliance in AI girlfriends and emotional dependence, and at where an AI companion ends and professional help begins in therapist vs. AI companion.

If you are struggling or thinking about self-harm, please contact a local crisis line or emergency services, or reach out to someone you trust.

Platforms and user numbers

Most platform numbers come from the companies themselves, and they count different things: monthly active users, daily active users, weekly users or cumulative sign-ups. Compare them with care.

Company-reported figures are not audited. MiniMax warns that one person may hold several accounts. The Replika figure is a homepage counter the company labels "users worldwide"; we read it as cumulative sign-ups, not active users.

Usage and rankings

• About 10% of Character.AI's roughly 20 million monthly active users were under 18 in October 2025, a share CEO Karandeep Anand said had been falling as the app shifted toward storytelling and roleplay. Company figures given to CNBC, not audited. The under-18 share depends on the ages users declare. Source: CNBC, 2025
• 80 minutes a day is the average time Character.AI users spend chatting with its AI characters, according to CEO Karandeep Anand, as reported in August 2025. A second-hand CEO claim. The report does not link the original interview or say whether the average is per daily or monthly user. Source: The Decoder, 2025
• 90 minutes is the average active session length Chai reported in July 2025, up from 60 minutes when it raised its previous round in January 2024, while its daily active users grew from 800,000 to 2 million over the same period. Self-reported in the company's own funding announcement. Source: Chai Research, 2025
• About 705 million hours were spent by U.S. users on AI companion apps in the first quarter of 2026, against about 280 million hours on dating and social-discovery apps. In the first quarter of 2025 the figures were roughly 580 million and 330 million hours. Treat as approximate: the numbers come from a small secondary blog that appears to read them off a Sensor Tower chart, and they could not be checked against Sensor Tower's gated report. They measure time spent, not number of users. Source: OfficeChai, reporting Sensor Tower, 2026
• Nearly 40% was the jump in Grok's mobile usage in July 2025, which a16z ties to the release of Grok 4 on July 9. xAI added AI companion avatars on July 14, and a16z says the anime avatar Ani was particularly popular at launch. xAI has not published usage figures for the companions themselves. The same report puts Grok at upwards of 20 million mobile monthly active users in total (Sensor Tower data). Source: a16z, 2025
• 1 of 14 products that appeared on all five editions of a16z's list of the top 50 generative-AI web products (September 2023 to August 2025) is Character.AI, which a16z files under companionship. The list ranks products by unique monthly web visits (SimilarWeb data). Source: a16z, 2025
• 7 companion sites appear on the web top 50 of a16z's sixth edition (January 2026 data): Candy Network, Character.ai, CrushOn, Janitor AI, Ourdream.ai, Polybuzz and Spicychat. Of these, only Character.ai and Polybuzz also appear on the mobile list. a16z publishes an alphabetical list; grouping these as companion sites is our own classification. Source: a16z, 2026
• About four hours a day is the average time Zeta users in Japan spend in the app, and in April 2026 the Korean character-chat app overtook Netflix for first place by total usage time in Japan's entertainment category. Industry figures reported by the newspaper without a named primary source. Source: Seoul Economic Daily, 2026

This page does not rank apps. For recommendations, see our ranking of the best AI girlfriend apps; for head-to-head matchups, our AI girlfriend app comparison hub; and for the platform that appears most often in these numbers, our Character.AI roleplay review.

Consumer spending and downloads

App-store data gives the clearest view of what people actually spend on AI companion apps. It covers mobile in-app purchases only, not web subscriptions or advertising. Our explainer on how AI companion apps make money covers the business models behind these numbers.

The app category

• $82 million was spent by consumers on AI companion apps worldwide in the first half of 2025, and the category was on track for more than $120 million for the full year. Appfigures estimate covering App Store and Google Play spending only, excluding web subscriptions and ads. We could not find a published full-year 2025 figure. Source: TechCrunch, Appfigures data, 2025
• $221 million in lifetime consumer spending had gone to AI companion apps worldwide by July 2025, and revenue so far in 2025 was 64% higher than in the same period of 2024. Appfigures estimate, mobile in-app spending only. Source: TechCrunch, Appfigures data, 2025
• 220 million downloads of AI companion apps had been recorded worldwide by July 2025, including 60 million in the first half of 2025 alone, up 88% year over year. Appfigures estimate. Source: TechCrunch, Appfigures data, 2025
• 337 active, revenue-generating AI companion apps were available worldwide as of mid-2025, and 128 of them had been released in 2025. Excludes apps that launched and were later removed. "Companion" follows Appfigures' definition: apps for chatting with premade or user-generated characters. Source: TechCrunch, Appfigures data, 2025
• 89% of AI companion app revenue goes to the top 10% of apps, and only about 33 apps had passed $1 million in lifetime consumer spending. Appfigures estimate as of July 2025. Source: TechCrunch, Appfigures data, 2025
• $1.18 was the revenue per download for AI companion apps in 2025 (January to July), more than double the $0.52 of 2024. Source: TechCrunch, Appfigures data, 2025
• 17% of active AI companion apps have "girlfriend" in their name, compared with 4% for "boyfriend" or "fantasy", making it the most common theme. This counts words in app names, not users or revenue. Source: TechCrunch, Appfigures data, 2025
• About 30% was the quarter-over-quarter rise in worldwide in-app purchase revenue for AI companion apps in the first quarter of 2026, the fastest growth of any generative-AI subcategory, according to Sensor Tower. Reported from Sensor Tower's gated APAC report; only the growth rate is public, not the dollar total. Source: Digital Today, reporting Sensor Tower, 2026
• Top two by absolute in-app revenue among AI companion apps worldwide were PolyBuzz and Chai, with Korea's Zeta third and LoveyDovey seventh. Secondary account of a gated Sensor Tower report. The period is implied to be the first quarter of 2026, and the order between PolyBuzz and Chai and the dollar amounts are not given. Source: Digital Today, reporting Sensor Tower, 2026
• $62 million in consumer spending went to the companion segment among the 1,000 top-earning AI apps from January 2023 to December 2024, ranking it sixth among AI segments by revenue and fourth by downloads. Older baseline data. The same report puts companion revenue growth at +2,990% in 2024 and says 63% of companion-app users are aged 18 to 24. Appfigures classified apps differently here, so this figure should not be compared directly with the $221 million above. Source: Appfigures, 2025

Company revenue

• $50 million is the annual revenue run rate Character.AI expected to reach by the end of 2025, according to CEO Karandeep Anand. The company makes money mainly from advertising and a $10 monthly subscription. A forward-looking CEO statement, not audited. We found no confirmed final 2025 revenue. Source: CNBC, 2025
• $40 million in annualized revenue was reported by Chai in July 2025, double the $20 million of January 2024, with a team of 11 engineers. Self-reported and unaudited. Source: Chai Research, 2025
• Over $100 million a year in revenue was claimed by Chai in a July 2026 press release, whose headline calls it "$100M ARR". A company-issued press release; self-reported and unaudited. Source: Chai via PR Newswire, 2026
• US$18.75 million in revenue came from MiniMax's companion apps Talkie and Xingye in the first nine months of 2025, 35.1% of the company's total, compared with US$19.46 million, or 63.7%, for full-year 2024. Talkie and Xingye had 1.39 million paying users in the first nine months of 2025. Nine-month figures are unaudited. Source: MiniMax prospectus, 2025
• About 26 billion won in 2025 revenue and about 3 billion won in operating profit were recorded by Scatter Lab, maker of the character-chat app Zeta. A newspaper report citing industry sources; no filing is cited and the figures are approximate. Source: Seoul Economic Daily, 2026

Investment and market-size estimates

Deals and funding

• $2.7 billion in cash was paid by Alphabet to Character.AI under an August 2024 agreement that gave Google a non-exclusive license to Character.AI's current large language model technology. Alphabet also canceled its convertible instruments, hired certain Character.AI employees and recorded $2.7 billion of goodwill and $413 million of intangible assets. Alphabet accounted for the deal as a business combination but did not describe it as an acquisition. Source: Alphabet Form 10-Q, 2024
• $150 million was raised by Character.AI in March 2023 at a $1 billion valuation, in a round led by Andreessen Horowitz, when the company was reportedly pre-revenue. The last officially announced priced round we found for Character.AI. Source: CNBC, 2023
• Over $55 million in total funding was reported by Chai after a 2025 strategic round, with CoreWeave and AMD among its backers. Self-reported; no round sizes or valuation disclosed. Source: Chai Research, 2025
• HK$4.8 billion (about US$620 million) was raised in the Hong Kong IPO of MiniMax, parent of the companion app Talkie, and its shares closed 109% above the offer price on their January 9, 2026, debut (HK$345 against an offer price of HK$165). MiniMax is a broad AI-model company; Talkie and Xingye made up 35.1% of its revenue in the first nine months of 2025. Source: CNBC, 2026
• About 50 billion won in investment was secured by Scatter Lab, maker of Zeta, from venture capital firms including Atinum Investment, SBVA and Mirae Asset Venture Investment, according to a June 2026 report. Based on industry sources. The report does not say when the round closed, and no valuation was reported. Source: Seoul Economic Daily, 2026
• $250 million was raised in an October 2025 Series B by Sesame, a conversational AI startup and smart-glasses maker that says its planned glasses will give access to an AI companion. Sesame is voice AI plus hardware rather than a pure companion app, and no valuation was stated. Its demo voices were accessed by more than a million people in the first few weeks, according to investor Sequoia. Source: TechCrunch, 2025

Market-size estimates

Research vendors publish "AI companion market" forecasts that are far larger than measured app spending, because they define the market much more broadly. We list them for completeness and flag them as vendor estimates. For a narrative overview of the market, see our article on the AI girlfriend market in the U.S. and worldwide.

• US$36.79 billion is Grand View Research's estimate of the global "AI companion market" in 2025, which it forecasts to reach US$317.96 billion by 2033, a compound annual growth rate of 31.0%. A vendor estimate with an unpublished methodology and a broad definition that includes enterprise and virtual-assistant uses. It is not comparable with consumer spending on companion apps. Cited from a reseller's report summary. Source: Grand View Research via GII, 2026
• US$37.12 billion is Precedence Research's estimate of the AI companion market in 2025, rising to US$48.63 billion in 2026 and about US$552.49 billion by 2035, a compound annual growth rate of 31.00%. A vendor estimate with a proprietary methodology and a broad definition covering mental health, education, personal assistance and business uses. It should not be read as consumer spending on companion apps. Source: Precedence Research, 2025

Lawsuits and regulation

Regulators and courts have focused mainly on minors: age checks, break reminders, crisis referrals and harmful conversations. Our AI companion laws tracker follows each law in detail, and our guide to AI girlfriends and U.S. law explains what the rules mean for users.

United States

• 7 companies received FTC orders on September 11, 2025, asking how they measure and limit harms to children and teens from AI chatbots that act as companions: Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap and xAI. The Commission voted 3-0. A Section 6(b) study, not a law-enforcement action. We found no published FTC findings from it as of September 2026. Source: U.S. Federal Trade Commission, 2025
• Up to $15,000 per day in civil penalties for a violation can be sought by New York's attorney general under the state's AI Companion Models law (General Business Law Article 47), which took effect on November 5, 2025. Summary from a law firm; the statute text is also online. Source: Morrison Foerster, 2025; statute text
• 33-3 and 59-1 were the Senate and Assembly votes for California SB 243, which Governor Newsom signed on October 13, 2025, and which took effect on January 1, 2026. Its author billed it as the first-in-the-nation set of safeguards for AI companion chatbots, but New York's AI companion law took effect earlier, on November 5, 2025. Source: Office of Sen. Steve Padilla, 2025
• $1,000 per violation, or actual damages if greater, is what a person injured by a violation of SB 243 can recover in a civil action, along with injunctive relief and reasonable attorney's fees and costs. The law also requires break reminders at least every three hours for users known to be minors, and from July 1, 2027 operators must report annually to California's Office of Suicide Prevention. Source: California Legislative Information, 2025
• January 1, 2027 is when the AI companion-chatbot laws Oregon (SB 1546) and Washington (HB 2225) enacted in March 2026 take effect. Washington requires an AI disclosure every hour, rather than every three hours, for users known to be minors or when the chatbot is directed to minors. Law-firm summary. Source: Mayer Brown, 2026
• 12 states had enacted companion-chatbot laws through the first half of 2026, according to MultiState, counting the 2025 laws in California and New York. Its own list, however, names only 11: Colorado, Georgia, Iowa, New York, Rhode Island, California, Idaho, Nebraska, Oregon, Connecticut and Washington. Check the list before citing the count. At the time, a Hawaii bill had been sent to the governor and Arizona's governor had vetoed a similar bill. Source: MultiState, 2026
• 44 attorneys general from 42 states plus American Samoa and the Northern Mariana Islands warned AI companies on August 25, 2025, to stop chatbots from engaging in sexually inappropriate conversations with children. Recipients included companion-app makers Chai AI, Luka Inc./Replika and Nomi AI, alongside Anthropic, Apple, Google, Meta, Microsoft, OpenAI, Perplexity AI and xAI. A warning letter, not a legal action. Source: Arizona Attorney General, 2025
• 5 lawsuits were the subject of settlement agreements reached in January 2026 by Character.AI, its founders and Google: Megan Garcia's wrongful-death suit over her son's suicide, filed in October 2024, and four other family cases in New York, Colorado and Texas. CNN reported the agreements from court filings; terms were not disclosed. Source: CNN, 2026
• January 2026: Kentucky's attorney general sued Character Technologies and its owners under the state's Consumer Protection Act and Consumer Data Protection Act, alleging a failure to use age gates, parental consent or identity verification and concealment of how children's data is collected and used. Law-firm summary; the attorney general's own release could not be opened. Source: Frankfurt Kurnit Klein & Selz, 2026
• May 1, 2026: Pennsylvania's Department of State filed a petition in Commonwealth Court (220 MD 2026) alleging that Character.AI chatbots practiced medicine without a license; one character presenting as a psychiatrist claimed to be licensed in Pennsylvania and gave an invalid license number. The governor announced the suit on May 5, 2026. The state is seeking a preliminary injunction. Source: Commonwealth of Pennsylvania, 2026

Outside the United States

• €5 million was the fine Italy's data protection authority imposed on Luka Inc., Replika's developer, in a final decision dated April 10, 2025. It found that until February 2, 2023, Luka had not identified a legal basis for its processing and had no age verification, and that its current age verification is still deficient. The authority reserved a separate proceeding on how the underlying generative AI system was trained. Source: European Data Protection Board, 2025
• Up to A$49.5 million in penalties can follow a failure to respond to the legal notices Australia's eSafety Commissioner sent in October 2025 to four AI companion providers: Character.AI, Glimpse.AI (Nomi), Chai and Chub AI. Law-firm summary. Source: MinterEllison, 2025
• 0 of 4 AI companion providers examined by eSafety had robust age verification, and three of them (Chai, Chub AI and Nomi) did not direct users to crisis support when self-harm was detected in user prompts, eSafety reported in March 2026. After the notices, and before the report, Chub AI geo-blocked its service in Australia. Cited from eSafety's media release on the Medianet newswire. Source: eSafety Commissioner, 2026
• 2 hours of continuous interaction triggers a mandatory usage pop-up under China's Interim Measures for AI anthropomorphic interactive services, issued on April 10, 2026, by the Cyberspace Administration of China and four other departments and in effect since July 15, 2026. Law-firm summary of China's first departmental regulation dedicated to these services. It also requires a minor mode and the removal of non-compliant products from app stores. Source: Bird & Bird, 2026

Platform changes

• 2 hours a day was the starting cap on open-ended chat for users under 18 that Character.AI announced on October 29, 2025, ramping down until the feature was due to end no later than November 25, 2025. It also said it would roll out age assurance using an in-house model and third-party tools including Persona. Source: Character.AI, 2025
• 1 hour a day was the U.S. cap for users under 18 by November 21, 2025, when Character.AI said removal of open-ended chat would start in the U.S. on November 24 in two stages, with other markets to follow. A second group of under-18 users was to lose open-ended chat weeks after the first, which took the change past the original November 25 date for some minors. Source: Character.AI, 2025

What the numbers mean

1. The definition drives the headline number. "Has ever chatted with a romantic AI", "uses a chatbot for companionship" and "has an AI friend" are different questions, and they produce very different answers. Questions about specific, ongoing uses, such as companionship or having an AI friend, give the lowest figures, and broad "ever used" questions the highest. When you cite a figure, cite its population and question with it.
2. Young people lead, and many use general chatbots. Figures for teens and young adults are consistently higher than those for all adults, and several of the largest studies count ChatGPT and similar assistants as companions. Studies of dedicated companion apps and studies of general chatbots measure overlapping but different behavior.
3. Measured spending is modest and concentrated. App-store spending on companion apps is small next to the vendor forecasts, and most of it goes to a small group of apps. The "AI companion market" estimates count a far broader set of products, so they should not be set side by side with app spending.
4. Heavy use is the consistent warning sign, but cause is rarely shown. Several U.S. studies link heavier or more intense use with loneliness, dependence or lower well-being, while a large Japanese study found AI companion use associated with higher well-being, most strongly among lonely adults. Most of this research is cross-sectional or not yet peer-reviewed, so it says more about who uses AI companions heavily than about what the apps do to them.
5. Regulation has moved from warnings to rules, focused on minors. In 2025 and 2026 the response shifted from warning letters and study orders to enforceable rules: state laws with break reminders and damages, lawsuits and settlements, regulator notices in Australia and national rules in China.

Methodology and sources

This page is desk research. We collected published statistics from surveys, peer-reviewed studies, preprints and working papers, company filings and announcements, regulator releases and news reports. We did not run our own surveys, interviews or product tests for this page. This edition lists 104 data points from 68 source documents.

Every statistic was re-checked against its source on September 19, 2026. Where checking showed that a figure needed a narrower description, for example a pooled survey result or the base of a subgroup, we use the narrower wording.

What we include, and how:

• Latest data first. We use the most recent figure available for each question and keep older figures only as baselines, labeled with their year.
• Primary sources preferred. We link the original report, filing or release wherever we could open it. Where only news or law-firm coverage was accessible, we name the original publisher and say that the source is secondary.
• Company and vendor figures are flagged. Self-reported user counts and revenue are labeled as company figures, and market-size estimates from research vendors are kept apart from measured consumer spending.
• Population and year stay attached. Every figure carries who was asked, when, and the sample size where it was published. We do not combine, re-weight or average numbers from different surveys.
• Caveats are kept. Known limitations, such as non-probability samples, preprints or funding by an interested company, are stated next to the number.

This page contains no affiliate links, and no company paid to be included. We review it quarterly and add new studies as they are published; the date at the top shows the last full review. If you spot an error or know of a newer source, contact us through our About page. Our editorial policy explains how we fact-check and correct articles.

<a href="https://aigirlfriendworld.com/ai-companion-statistics">AI Companion Statistics 2026</a> (AI Girlfriend World)

More research from AI Girlfriend World

• AI companion laws tracker: laws and regulations that apply to AI companion apps.
• AI companion privacy scorecard: what the privacy policies of AI companion apps say about your chats.
• AI companion history timeline: how AI companions developed over time.

Related guides

• The AI girlfriend market in the U.S. and worldwide
• What is an AI girlfriend?
• AI companions and loneliness
• AI girlfriends and emotional dependence
• AI girlfriends and U.S. law