Digital democracy|Digital privacy
Love in the online age: the growth of AI companions and their privacy issues
The online world is experiencing a significant transformation as millions of people are now forming emotional and romantic relationships with AI companions. With the rapid growth of platforms like ChatGPT, a vast global market for AI companions has emerged, and it is expected to grow fivefold over the next decade. Our in-depth study explores this swift rise in popularity, analyzing major platforms to assess their true reach and, more importantly, to uncover hidden privacy risks and data-collection practices that could expose users' most intimate secrets.
Global rise of AI companions
In early 2023, catalyzed by the popularity of ChatGPT, numerous AI companions were launched, allowing people to have conversations with avatars of famous individuals and popular fictional characters. This led to people forming emotional and romantic connections with AI. Currently, millions of people worldwide use AI companions daily, and the market is projected to grow from a $50B industry in 2026 to $436B over the next decade¹.
To understand the scale of adoption, we used Ahrefs to analyze 10 leading AI companions and found that their websites collectively attract 14M monthly organic search visits. It is important to note that total web traffic is likely much higher, as these figures exclude direct visits and traffic originating from within companion apps. Character AI and PolyBuzz had the most visits, with over 6.4M and 6.2M, respectively, followed by Talkie with 706k, CHAI with 221k, Love and Deepspace with 182k, Zeta with 104k, Replika with 81k, Nomi with 48k, and Kindroid with 32k.
Looking at the male-to-female ratio, Similarweb data shows that Character AI has an equal number of male and female users, each accounting for about 50% of the traffic. PolyBuzz has 63% male and 37% female users, while Talkie has 62% male and 38% female users. Combining these numbers and considering all of the traffic to these three most popular AI companion platforms, we find that, on average, males use these AI companions 56.5% of the time, while females use them 43.5% of the time.
In terms of actual users, Character AI has around 20M monthly active users and reportedly receives approximately 181M monthly visitors². Not all platforms report such data, but CHAI has released a report indicating it has 10M users³. Zeta is reported to have 6M users⁴, and Love and Deepspace made a Facebook post claiming that they have reached 50M global users⁵. Beyond website visits, the popularity of AI companions is also evident in mobile app adoption. While the Paradot AI platform sees fewer website visits, its mobile app has still achieved over 10M downloads across the Play Store and App Store. Meanwhile, the most popular AI companion apps, Character AI and PolyBuzz, each have over 50M downloads. This widespread adoption of AI companions across platforms demonstrates a huge and growing demand, which is expected to increase even further as more people embrace AI companionship.
Where are AI companions gaining the most traction?
Organic search data reveals that AI companion sites attract visitors from 142 countries, with the US leading at 33.5% of visits. This is followed by Brazil (8.7%), Indonesia (6.9%), Mexico (4.1%), the UK (3.8%), India (3.4%), Canada (2.9%), France (2.2%), and the Philippines (1.9%). Together, these nine markets account for 67.5% of visits, while the remaining 32.5% are distributed across the rest of the world.
While total organic search traffic highlights the most significant sources of AI companion users, adjusting for population size reveals which countries are truly embracing AI companionship. Interestingly, the top eight countries come from five different continents. Leading the list is the US, with 136 per 10,000 people reaching AI companion websites via organic search, followed by its neighbor, Canada, with 102. The following three countries on the list are all from different continents: Chile (90), Australia (83), and the UK (78). Singapore rounds out the top eight with 72, making it the highest-ranking Asian country.
A closer look at organic search traffic — adjusted per 10,000 people — reveals the highest AI companion usage rates across continents, along with the total traffic for each continent.
- North America: with a total traffic of 5.9M, the highest usage rates are seen in the US (136), Canada (102), Puerto Rico (52), El Salvador (51), and Costa Rica (47);
- Asia: with a total traffic of 3M, the highest usage rates are seen in Singapore (62), Malaysia (49), the UAE (42), Bahrain (39), and Indonesia (34);
- Europe: with a total traffic of 2.5M, the highest usage rates are seen in the UK (78), France (45), Finland (45), Latvia (44), and Spain (43);
- South America: with a total traffic of 2.2M, the highest usage rates are seen in Chile (90), Uruguay (76), Brazil (58), Peru (56), and Argentina (46);
- Oceania: with a total traffic of 0.26M, the highest usage rates are seen in Australia (83) and New Zealand (70);
- Africa: with a total traffic of 0.19M, the highest usage rates are seen in Mauritius (23), South Africa (11), Algeria (6), Libya (6), and Tunisia (5).
For a complete breakdown of usage by country, please refer to the visual below.
How data-hungry are AI companions?
To provide meaningful interactions and personalize experiences, AI companions need to store chat history, which can serve as training material for AI models. For example, Character AI’s privacy policy⁶ explicitly states that user data is utilized to train its AI and machine learning models. The platform strongly advises users, ‘’Please do not include any sensitive personal information in your interactions with us or the Services.’’ Furthermore, these platforms typically do not provide full end-to-end encryption for conversations, meaning private chats could potentially be exposed in a data leak.
To better understand what additional data AI companions collect, we analyzed their App Privacy Labels on the Apple App Store. Character AI stands out, collecting 18 out of 35 possible unique data types. These include user and device IDs, email address, usage data (such as product interactions, advertising data, and search history), purchase history, location, and user-generated content such as gameplay, photos, videos, and audio, potentially enabling detailed user profiling. In comparison, the least data-hungry apps are PolyBuzz, which collects only the device ID, and Talkie, which collects both device and user IDs.
Our analysis also found that Character AI, Zeta, Love and Deepspace, and CHAI all collect users’ photos or videos. Additionally, both Character AI and Love and Deepspace gather audio data, such as voice recordings. Collecting these types of media content raises privacy concerns, as images, videos, and audio clips are difficult to anonymize and are at risk in the event of a data leak. As more people integrate AI into everyday life, users should be especially mindful of the media they share, understanding that these materials may be stored, analyzed, and potentially repurposed by the platform.
Purposes behind data collection by AI companions
Ideally, apps should collect only the data essential for functionality, with limited analytics. However, these platforms often gather more than necessary. Every app studied collects data for analytics to evaluate user behavior and features. Character AI leads this trend, collecting 14 data types for analytics alone. In comparison, Love and Deepspace, Nomi, and Paradot each collect 7; Zeta collects 5; CHAI collects 4; Replika and Kindroid each collect 2; and Talkie and PolyBuzz collect only 1 analytics data type each.
The most sensitive practice identified is the use of data for tracking purposes. This data could be sold to data brokers or used to display targeted advertisements in your app. Nine out of 10 apps collect such data, with Character AI leading by collecting your contact information, identifiers, location, and usage data. Meanwhile, Nomi doesn’t collect data for tracking purposes. Other apps collect either identifiers or usage data for tracking, except for Replika, which also collects contact information.
Personalization, the process of customizing user experiences like recommended posts or suggestions, is also a primary focus for these platforms. This data helps apps learn about your interests and habits. Eight out of 10 apps collect data for this purpose, again led by Character AI with 11 types and Paradot with 8 data types. Zeta, Love and Deepspace, and Nomi each collect 3 data types, while Replika, Kindroid, and Talkie each collect 1.
Methodology and sources
For this study, we selected 10 popular AI companion platforms and analyzed their average monthly organic search traffic in each available country using Ahrefs. Data collection was conducted on January 16, 2026, covering the period starting from January 1, 2026. We then aggregated total organic search traffic across all 10 AI companion websites, excluding records with fewer than 100 organic visits. To gain clearer insights into which countries have the highest relative traffic, we normalized traffic per 10,000 people and excluded countries with populations below 1M to gain clearer insights into which countries have the highest relative usage.
In the second part of the study, we analyzed the privacy disclosures on the Apple App Store. We examined the number of data types each AI companion collects and the stated purposes for collecting this data.
Lastly, to be included in this study, AI companions were required to have both a publicly available website and a corresponding app listed in the App Store.
For complete research material behind this study, visit here.
