The AI companion category has moved from novelty to measurable industry in under three years. What makes it unusually difficult to write about is that the numbers describing it disagree with each other by an order of magnitude — not because anyone is wrong, but because different sources are measuring different things and rarely say which.
This piece separates those measurements. Market forecasts, observed consumer spending, download figures, user demographics, and the research on psychological effects are all covered separately, with sources named throughout and the disagreements shown rather than resolved.
Market Size: Why the Numbers Range From $2 Billion to $48 Billion
The single most important thing to understand about AI companion market data is that "AI companion" means different things to different research firms.
Broad market definitions (voice assistants, multimodal AI, enterprise applications, consumer companions):
- Grand View Research valued the global AI companion market at $36.8 billion in 2025, estimating $48.0 billion for 2026, with projections reaching $318 billion by 2033 at a 31.0% CAGR
- The Business Research Company sized the market at $18.35 billion in 2025, growing to $24.09 billion in 2026 at 31.3% CAGR, reaching $70.91 billion by 2030
- Fortune Business Insights valued it at $37.73 billion in 2025, projecting $49.52 billion in 2026 and $435.9 billion by 2034
- Research and Markets sized the AI girlfriend app segment specifically at $2.32 billion in 2025 and $2.91 billion in 2026, roughly 25.5% CAGR
- Appfigures data reported by TechCrunch put consumer spending in dedicated companion apps on track for approximately $120 million in 2025
Downloads and Active Users
Download and engagement figures come from app-store intelligence firms and are among the harder numbers in this category.
- AI companion apps passed approximately 220 million cumulative downloads by mid-2025
- First-half 2025 downloads were up 88% year over year
- 337 active revenue-generating AI companion apps were operating worldwide as of 2026, with 128 released in 2026 alone
- Character.AI reports approximately 20 million monthly active users
- Replika reports 40 million-plus registered users (a company-stated figure rather than independently verified)
- One estimate placed active AI companion app users at approximately 50 million by February 2026, generating $500 million-plus in annual subscription revenue
Who Actually Uses AI Companions
Demographic data is the area where the category most consistently defies assumptions.
Teenagers:
- 72% of US teens have used an AI companion
- 52% use one regularly
- 13% use one daily (Common Sense Media)
- Nearly 1 in 5 US adults have chatted with a romantic AI
- That rises to approximately 1 in 3 young men and 1 in 4 young women
- Users aged 18–24 make up over 65% of the global audience
- 19.1% of American adolescents and young adults have used AI chatbots for emotional or psychological advice, up from an earlier figure of one in eight (JAMA Pediatrics)
- More than 40% of those users consult chatbots at least monthly
- ChatGPT's own figures suggest nearly one million people per day use the service for emotional support, mental health, or self-harm concerns
- Entertainment: 30%
- Curiosity about the technology: 28%
- Advice: 18%
- "The Lonely" — approximately 40%
- "The Anxious," practising social skills — approximately 25%
- "The Augmenters," supplementing existing relationships — approximately 12%
- "The Bereaved," processing grief — approximately 8%
The demand side of this market is inseparable from documented loneliness trends.
- 67–80% of Gen Z report recent loneliness across global surveys, despite being the most digitally connected generation on record (GWI)
- Young men aged 15–29 are the loneliest demographic in the Western world — 25% report feeling lonely versus 17% for all other adults
- 44% of young adults aged 18–25 report feeling they "don't matter to others"
- The World Health Organization identifies social isolation and loneliness as critical health concerns affecting all age demographics
This is where the evidence genuinely divides, and any statistics resource that presents only one side is misrepresenting the state of the research.
Findings supporting benefit:
- Research published in the Journal of Consumer Research (De Freitas et al., Harvard/Wharton) found AI companions reduce loneliness about as much as interacting with another person. The mechanism the researchers isolated was feeling heard — attention, empathy and respect rather than humanity
- A JMIR randomized trial using the Luda Lee chatbot found university students with elevated loneliness showed statistically significant loneliness reduction by week two and social anxiety reduction by week four
- South Korean researchers reported regular AI social chatbot interaction reduced loneliness by an average of 15% and social anxiety by 18% among student participants
- A study of 14,721 Japanese adults found AI companion use significantly associated with higher scores across evaluative, hedonic and eudaimonic well-being domains — while non-companion AI use showed weaker or inconsistent associations
- Maples et al. (npj Mental Health Research, 2024) surveyed 1,000+ student Replika users and found 3% reported the app had halted their suicidal ideation
Findings indicating harm:
- A two-year Aalto University study of nearly 2,000 Replika users found heavy use quietly raised the perceived cost of human relationships and correlated with rising signals of depression, loneliness and suicidal ideation (CHI 2026). Lead researcher Talayeh Aledavood described the finding as a paradox: unconditional support attracts people struggling socially while making messy, effortful human relationships feel more costly by comparison
- Stanford research published in Nature Human Behavior found opening up to chatbots about personal issues risks a cycle where people with limited social networks are drawn to intense AI use that further reduces real-world contact
- A 2026 controlled experiment compared two weeks of daily chatbot use against two weeks of human peer conversation. Chatbots produced no lasting improvement in loneliness. Only human interaction did
- The OpenAI/MIT Media Lab affective use study found heavy ChatGPT users showed lower socialisation and increased emotional dependence over time
- A 12-month longitudinal study found heavier ongoing companion use linked to increased loneliness over time
The pattern across both sets of findings: short-term relief is well-evidenced. Long-term outcomes depend heavily on whether use supplements human connection or substitutes for it. Studies measuring hours or weeks tend to find benefit. Studies measuring months or years tend to find risk.
Regulation
Legislative response accelerated significantly in 2026.
- California SB 243 took effect January 2026, requiring companion apps to clearly disclose that users are talking to a bot and to refer users expressing suicidal thoughts to crisis services
- Twelve US states had enacted companion-chatbot legislation by mid-2026, several writing content rules directly into law
- Italy's data protection authority fined Replika operator Luka €5 million in May 2025 for data-protection violations flagged in February 2023, including absent legal basis for processing and inadequate age verification. A separate inquiry into the platform's model training data remains open
- The World Health Organization published guidance on responsible AI for mental health in 2026
Platform Concentration and Retention
- Character.AI holds the largest active user base in dedicated companion apps at approximately 20 million MAU
- Text-based AI companions held the largest market share by category in 2025
- North America leads regionally with revenue share estimates ranging from 33.5% to 41% depending on source, followed by Asia-Pacific at approximately 32%
- When Replika removed erotic roleplay in February 2023, research analysing the community response found 85% of affected users had formed emotional attachments to their companions, with a documented share describing the loss in the language of grief
Quick Reference Table
| Metric | Figure | Source |
|---|---|---|
| Broad market size 2026 | $24.09B – $49.52B | The Business Research Company / Fortune Business Insights |
| Companion app segment 2026 | $2.91B | Research and Markets |
| Observed app spending 2025 | ~$120M | Appfigures / TechCrunch |
| Cumulative downloads | ~220M (mid-2025) | App-store intelligence |
| YoY download growth H1 2025 | +88% | App-store intelligence |
| Active companion apps | 337 (128 launched 2026) | Industry tracking |
| US teens who have used one | 72% | Common Sense Media |
| US adults using romantic AI | ~1 in 5 | Survey data |
| Users aged 18–24 | 65%+ of global audience | Industry analysis |
| Gen Z reporting loneliness | 67–80% | GWI |
How to Read These Numbers
Three cautions apply to every figure on this page.
Definitional variance is the largest source of disagreement. A forecast covering enterprise voice assistants and one covering AI girlfriend apps will differ by an order of magnitude while both being accurate. Always check what a market figure includes before comparing it to another.
Company-stated user numbers are not audited. Registered users, monthly active users and downloads measure different things, and platforms tend to cite whichever is largest.
The psychological research is early and contested. Most studies are small, self-reported, and short-duration. The longer-running studies find different results from the shorter ones, and the field has not resolved whether that reflects genuine long-term harm or selection effects in who becomes a heavy user.
Figures are current as of publication. Where sources disagree, ranges are shown rather than a single selected figure.