Digital democracy|Digital privacy
AI companion apps “love” your personal data
Love can be expressed in countless ways — through loved ones, family, pets, possessions, or even virtual beings. As AI technology advances, companies are exploring new ways to enhance human connections, leading to the rise of innovative solutions like AI companions. What was once considered science fiction is now becoming a reality, with more people embracing digital friendships and even digital partnerships. This evolution raises intriguing questions about the future of human connections and encourages us to explore how these new forms of attachment can be pursued safely in our daily lives.
Key insights
- AI companions are artificially intelligent systems designed to engage users in human-like interactions, offering emotional support, companionship, entertainment, and more.¹ Such services hold great promise in an era where loneliness is recognized as a global public health issue. According to the WHO, current global estimates suggest that 1 in 4 older adults experience social isolation, and between 5 and 15 percent of adolescents experience loneliness.² In response, many adults are turning to digital friendships and partnerships to combat these feelings of isolation. However, it's important for users to remember that these digital connections ultimately represent transactional relationships with the for-profit companies behind them.
- Within the domain of AI companionship, it appears that none of the AI leading companies are offering solutions for those who are interested in exploring digital relationships. Major players like OpenAI or Anthropic remain outside the spotlight, prompting questions about the reasons for their absence and the risks they might be trying to avoid. Nevertheless, users still have the potential to experiment with customizing AI chatbots on their platforms. This study focuses solely on apps that are competing for the top industry position as AI companions and are currently available in the App Store: Kindroid, Nomi, Replika, EVA, and Character AI.
- AI companion app developers can monetize users’ relationships through subscriptions and possibly through sharing user data for advertising.³ A review of the data collection practices of the analyzed apps indicates that 4 out of 5, or 80%, may use data to track their users. On average, these apps use two types of data for tracking. “Tracking” refers to linking user or device data collected from the app such as a user ID, device ID, or profile, with third-party data with user or device data collected from other apps, websites, or offline properties for targeted advertising purposes. Tracking also refers to sharing user or device data with data brokers.⁴ Nomi, 1 out of 5 analyzed apps, stands out by stating that it does not collect data for tracking purposes.
- Research shows that Character AI is the app that truly “loves” its users’ data. The analyzed apps typically gather about 9 out of 35 unique types of data. In contrast, Character AI shines through as it may potentially collect up to 15 types, nearly doubling the average. Alongside Character AI, another companionship app, EVA, emerges as the second most data-loving app, gathering 11 types of data. Both apps seek users' coarse location information, which is often leveraged in advertising to deliver targeted ads.
- Moreover, by analyzing user-provided content during conversations with AI companions, app developers can potentially access data that was previously out of reach. Given that users may form emotional connections with their AI companions and that these algorithms are designed to be nonjudgmental and available 24/7, people may disclose even more sensitive information than they would to another human being. This may lead to unprecedented consequences, particularly as AI regulations are just emerging.
Methodology and sources
For this study, the five AI companion apps, frequently mentioned by top media portals, were selected. Information on these apps’ data collection practices was sourced from the App Store on February 6, 2025. The App Store provides a list of 35 unique data types categorized into 16 categories.⁴ In this study, the data was analyzed based on the number, type, and handling of the data types collected by each app.
Note: For clarity, app names have been shortened in the visualization and the text above. The full names of these apps are available in the research material provided below.
For the complete research material behind this study, visit here.
