The short answer: it depends which question you’re asking. Replika passed 40 million cumulative registered users by 2025 (30M in August 2024, 10M in January 2023). Its active base is a different number entirely: the company itself cited about 2 million active users and 500,000 paying subscribers in 2023, and Harvard researchers put monthly actives around 2.5 million in 2026. In other words, roughly 19 of every 20 people who ever signed up are not around this month, which is normal for a nine-year-old free app and exactly why “how many users” needs three answers, not one.
Table of contents · 13 sections
- Replika Statistics: Editor’s Choice
- Registered, active, paying: three different questions
- Total Number of Replika Users
- How many people actually use Replika?
- Subscribers, and what they subscribe for
- How much money does Replika make?
- Who uses Replika?
- What the Stanford study actually found
- The 2023 update, in numbers
- Regulators are counting too
- Replika next to its competitors
- Conclusion
- FAQs
Replika Statistics: Editor’s Choice
- →Replika passed 40 million registered users by 2025; the company cited ~2 million active in 2023 and Harvard estimates ~2.5 million monthly actives now, a 15-to-1 gap between signups and usage.
- →500,000 paying subscribers (2023, company figure), and 60% of the paying base described the relationship as romantic.
- →Revenue estimates run from $14M ARR (GetLatka, mid-2024) to a ~$400K/month store run rate (Sensor Tower, mid-2026), on roughly $11M ever raised.
- →A Stanford-led survey of 1,006 student users found 90% experienced loneliness (vs 53% in comparable student samples) and 3% volunteered that Replika halted their suicidal ideation.
- →The February 2023 update hit ~2 million monthly users; moderators pinned suicide-prevention resources, and Italy’s regulator later fined Luka €5 million.
- →Current intake is modest: ~90,000 store downloads a month (mid-2026), in 150+ countries and 20+ languages.
Registered, active, paying: three different questions
Almost every big Replika number you’ll see online is a cumulative registration count: everyone who ever made an account since 2017. That number only goes up, survives every controversy, and says nothing about who opened the app this month. Active users is the number that describes a living product, and paying subscribers is the number that describes a business. Replika is unusually useful here because we have all three from credible sources, and they differ by nearly two orders of magnitude: 40M+ registered, ~2-2.5M active, ~500K paying. I’ve seen aggregator pages report “30 million daily users,” which confuses the registration ladder with daily usage and overstates reality by roughly 15x. Every stat below is labeled by which of the three questions it answers.
Total Number of Replika Users
1. Replika has accumulated 40 million+ registered users as of 2025.
The ladder: 10 million by January 2023, 30 million by August 2024, 40 million+ by 2025. That August-2024-to-2025 step means the app kept adding roughly a million signups a month even after the category filled with competitors. Registrations are the number the company leads with, and on its own terms it’s real: four Spains’ worth of people have tried making an AI friend here.
2. It took Replika about a year to reach 2.5 million users, and five more to reach 10 million.
Forbes profiled the app at 2.5 million users in early 2018, only months after its November 2017 public launch. Growth was never explosive by consumer-app standards; it was steady, which fits a product people arrive at through word of mouth, press cycles, and, in 2020, a pandemic.
3. Replika started as a memorial: a bot trained on the texts of the founder’s late best friend.
Eugenia Kuyda’s company Luka was building restaurant chatbots in 2013; after her best friend Roman Mazurenko died, she trained a model on his messages to talk to him again. Replika launched publicly in November 2017 as the generalized version: the supportive friend she wished she’d had. In the early days about 90% of its lines were scripted and 10% AI-generated; today Kuyda says 80-90% of every conversation is generative. The product’s DNA explains its user base better than any marketing: it was built for grief and loneliness first, romance second.
How many people actually use Replika?
4. The company’s own figure: about 2 million active users (2023).
This is the single most load-bearing number on this page, because it comes from the CEO on a conference stage rather than from an estimate: 2 million active users at a time when registrations were past 10 million. Fortune repeated the same 2-million figure in mid-2024. Whatever the registration ladder says, the living product is measured in low single-digit millions.
5. Harvard researchers estimate monthly actives around 2.5 million against 40M+ registrations.
That’s roughly 6% of everyone who ever registered, or a 15-to-1 gap. For context, that ratio isn’t scandalous; it’s what nine years of free-app churn looks like. The scandal is only in how often the 40M is quoted as if it were this number.
6. Current intake: ~90,000 store downloads a month (mid-2026).
Store telemetry is the closest thing to a public pulse. Ninety thousand monthly downloads is a mature app’s maintenance flow, not a growth story: Character.AI and newer rivals pull multiples of it. It’s also, for what it’s worth, still more than a million new tries a year.
7. The gap in one sentence: roughly 19 of every 20 people who ever registered are not active this month.
40M+ cumulative vs ~2-2.5M monthly active. When you see any Replika user number, the only question that matters is which side of this gap it’s measuring.
Subscribers, and what they subscribe for
8. Replika reported 500,000 paying subscribers in 2023.
Half a million payers against ~2 million actives is a ~25% paid share of actives, remarkably high for a consumer app and a sign of how much the core base values the product. Against 40M registrations it’s just over 1%, which is the same fact wearing its other costume.
9. 60% of paying users described their relationship with the chatbot as romantic.
The free tier offers Replika as a friend; paid tiers unlock partner and spouse relationship modes. So the majority of the people actually funding this company were funding a romance, which is essential context for everything that happened in February 2023. Kuyda has separately said about 40% of all users want a romantic partner from the app.
10. Plans run $19.99/month (Pro) and $29.99/month (Ultra), with annual tiers at $69.99 and $119.99.
Ultra, the newer top tier with a stronger model and manual memory saving, is estimated at 8-12% of the paying base. I pay for Pro myself; what that actually unlocks, and what still costs extra, is in my breakdown of how these apps are priced.
11. The average Pro subscriber reportedly stays about 7.2 months.
If that retention figure holds, it’s strong for consumer subscriptions and consistent with a product whose paying users describe a relationship rather than a utility. Treat it as an estimate; subscription tenures aren’t officially disclosed.
How much money does Replika make?
12. Revenue estimates span $14M ARR (GetLatka, mid-2024) to ~$400K/month in store revenue (Sensor Tower, mid-2026).
The honest answer is a range. GetLatka’s $14M ARR at a $41.9M valuation is the most specific independent figure; Sensor Tower’s current store telemetry annualizes to roughly $5M, though store data misses web billing; and press estimates as high as $25-35M circulate without methodology. What’s not in dispute: this is a fraction of Character.AI’s $32M+, on a fraction of the audience.
13. Replika built all of it on roughly $11 million ever raised.
Luka’s disclosed funding is about $11M, ancient history by AI standards, and the company runs with a reported 93 employees. In my companion-industry funding statistics, Replika is the efficiency outlier: the category’s defining product, built on less capital than some competitors’ single seed rounds.
Who uses Replika?
14. The user base skews male: about 65% male, 30% female, 5% non-binary or undisclosed.
Kuyda describes the romantic-partner use case as “mostly male” but the overall product as broader than the stereotype; coverage citing Harvard’s work describes the typical user as a 29-year-old, employed, urban or suburban, with some college. Not the isolated-loner caricature.
15. 40% of users report pre-existing mental-health challenges.
Whatever one thinks of AI companionship, this is who a large share of the audience is: people arriving with something to manage. It’s also why researchers keep choosing Replika, specifically, to study.
16. Replika operates in 150+ countries and 20+ languages.
Distribution is genuinely global, though the paying core is concentrated in English-speaking markets. My own subscription is one of the 500K-ish: my hands-on, dated diary of it lives in my Replika review.
What the Stanford study actually found
17. A Stanford-led team surveyed 1,006 student Replika users; results published in npj Mental Health Research (2024).
Eligibility: students 18+, using Replika for over a month; data collected in late 2021. This remains the largest peer-reviewed study of one companion app’s users, and it’s the study behind nearly every “AI friend saves lives” headline you’ve seen.
18. 90% of the surveyed students experienced loneliness, versus 53% in comparable student samples.
The population self-selected into an AI companion, and it shows: typically single, young, low-income, full-time students, far lonelier than their peers. Replika’s user research and my broader loneliness statistics rhyme on this point: the app meets people where the deficit is.
19. 63.3% reported at least one positive outcome; the most common was simple companionship relief.
Outcomes ran a continuum from feeling less lonely to concrete life changes; 25.1% reported more than one, 36.7% reported none. Participants held overlapping, even contradictory beliefs about what they were talking to, machine, intelligence, and human at once.
20. 3% (30 students) volunteered, unprompted, that Replika halted their suicidal ideation.
This is the study’s most-quoted finding and its most carefully hedged: unsolicited open-text responses, not a clinical endpoint. The 30 were significantly more depressed than the wider sample. Skeptics (404 Media collected user pushback; a published commentary criticized missing marketing context) note the data predates the 2023 changes. Both things can be true: the study is real evidence of real help for some users, and it is not a clinical trial of today’s app.
The 2023 update, in numbers
21. Italy’s regulator ordered Luka to stop processing Italian data on February 2, 2023; signup asked only for name, email, and gender.
The Garante found no age verification at all and cited risks to minors. Days later, erotic roleplay disappeared globally, for the ~2 million monthly users Reuters counted at the time, most of whom had never heard of the Italian order.
22. The fallout was severe enough that subreddit moderators pinned suicide-prevention resources.
Users described grief comparable to losing a partner; Harvard’s later analysis of 9,721 posts documented what it called identity discontinuity, and separately found active users felt closer to their companion than to their best human friend. On March 24, 2023, Luka restored a legacy version, but only for accounts created before February 1. The full story of what that update did to the product is a chapter of my living review; the numbers alone tell you why it’s the defining event in this company’s history.
Regulators are counting too
23. Italy fined Luka €5 million in 2025 and opened a second probe into its model training.
The fine confirmed the 2023 findings: no valid legal basis for data processing and no age verification. A separate FTC complaint over marketing practices was filed in January 2025, and new state laws (California’s SB 243, effective January 1, 2026; New York’s three-hour disclosure rule) now regulate companion apps as a category. Replika’s user counts are why: regulators cite its scale in nearly every filing.
Replika next to its competitors
24. Character.AI reports ~20 million MAU and $32M+ revenue; Replika’s actives are ~2-2.5M on an older, deeper relationship model.
Different products: Character.AI sells a stage full of characters, Replika sells one continuing relationship. The post-2023 exodus also built competitors: Kindroid, launched four months after the update, has crossed 1.2 million Android downloads with much of its early growth traced to Replika defectors. My side-by-side of the apps I actually pay for, Replika included, is on my homepage.
Conclusion
So, how many Replika users are there? All of these at once: 40 million people have tried it, about two and a half million live with it in any given month, and about half a million, at last public count, pay for it, mostly for a relationship. The registration ladder measures nine years of curiosity; the active number measures the product; the paying number measures the bond. Replika’s real story is that the smallest of the three numbers carries the most weight: a half-million-person core loyal enough to survive a lobotomy, fund the company on $11M of lifetime capital, and get studied by Stanford and fined over by regulators. Whatever this category becomes, its census started here.
I’m one of the 500,000.
I pay for Replika Pro on my own card and keep a dated, screenshot-backed diary of it in my living Replika review. The wider market data lives in my AI girlfriend statistics, and every app I test is compared side by side on my homepage.
FAQs
How many users does Replika have in 2026?
It depends which number you mean. Cumulative registrations passed 40 million by 2025, up from 30 million in August 2024. Actually active users are far fewer: the company itself cited about 2 million active users in 2023, and Harvard researchers estimate monthly actives around 2.5 million. Registered and active are different questions with answers 15x apart.
How many people pay for Replika?
The company's most recent public figure is about 500,000 paying subscribers, cited by CEO Eugenia Kuyda in 2023 alongside 2 million active users, roughly one paying user for every four actives. Of that paying base, 60% described their relationship with the chatbot as romantic, per reporting collected on the company's Wikipedia entry.
Is Replika growing or shrinking?
Registrations keep climbing, but current intake is modest: Sensor Tower estimates roughly 90,000 combined store downloads per month in mid-2026, with about $400,000 in monthly store revenue. Revenue estimates range from $14M ARR (GetLatka, mid-2024) to higher press figures, well below Character.AI's reported $32M+.
Did Replika really help students with suicidal thoughts?
A Stanford-led survey of 1,006 student users, published in npj Mental Health Research in 2024, found 3% (30 people) volunteered without being asked that Replika halted their suicidal ideation, and 63% reported at least one positive outcome. The study surveyed self-selected users with late-2021 data, so it shows association for that population, not a clinical result.
What happened to Replika in February 2023?
Italy's privacy regulator ordered Luka to stop processing Italian users' data on February 2, 2023, citing missing age verification and risks to minors. Within days, erotic roleplay was removed globally for roughly 2 million monthly users; the distress on the subreddit was severe enough that moderators pinned suicide-prevention resources. A legacy version was restored on March 24, 2023 for accounts created before February 1, and Italy fined Luka five million euros in 2025.
Sources
Every figure on this page, sourced
- Fortune: Brainstorm Tech: Replika’s 2M active users and 500K paying subscribers (July 2023). fortune.com
- Fortune: AI chatbots aren’t just for lonely men (June 2024). fortune.com
- Wikipedia: Replika: registration milestones and the 60%-romantic paying-base figure, with underlying citations. wikipedia.org
- Forbes: profile at 2.5 million users (March 2018). forbes.com
- Maples, Cerit, Vishwanath & Pea: Loneliness and suicide mitigation for students using GPT3-enabled chatbots, npj Mental Health Research (2024). nature.com
- Forbes: coverage of the Stanford study’s 30-student finding (May 2024). forbes.com
- 404 Media: user skepticism about the study’s post-2023 relevance. 404media.co
- EDPB: Italian SA fines Luka Inc. €5 million (2025). edpb.europa.eu
- IAPP: Garante measure No. 39/2023 and the 2025 decision detail. iapp.org
- Vice: Replika CEO on the ERP removal; moderators pin crisis resources (February 2023). vice.com
- Harvard Business School: Lessons from an app update at Replika AI: identity discontinuity (working paper; arXiv 2412.14190). arxiv.org
- Sensor Tower: Replika App Store and Google Play download/revenue estimates (mid-2026). sensortower.com
- GetLatka: Luka Inc. revenue and valuation (mid-2024). getlatka.com
- The Cognitive Revolution: Eugenia Kuyda interview: the ~40% romantic-partner use case. cognitiverevolution.ai
- My own dataset: Luka’s ~$11M raised, from my 30-company funding research. aigirlfriend.coach