Based on 60 in-depth U.S. interviews, this report maps five AI companion user segments, their hierarchy of needs, and the gaps brands can fill.
The U.S. Surgeon General has classified loneliness as a public health crisis. Every day, one in five adults feels lonely. At the same time, an aging inflection point is approaching: in 11 U.S. states, people aged 65 and over already outnumber children.
Technology is accelerating just as quickly. Multimodal capabilities reached maturity in rapid succession from 2024 to 2025. Voice, vision, memory, and emotion recognition—the four capabilities that underpin emotional relationships with AI—are all ready for practical deployment. Emotional companionship is moving from a niche experiment to a high-growth need, drawing bets from leading investors: Microsoft acquired Inflection's core team for $650 million; Google completed a $2.7 billion licensing deal with Character.AI; and Mattel partnered with OpenAI to develop a companion app.
Trooly.AI and Xiaguang Think Tank jointly released the *U.S. AI Companion Product User Insights Report*. Based on 1:1 in-depth AI video interviews with 60 real users in the United States, the report systematically unpacks the hierarchy of needs and user personas in AI companionship, exploring the kinds of AI products that can meet consumers' genuine emotional needs.
(The full report is 50 pages long. The following is an excerpt.)
The report uses the PEST framework to examine the external environment of the U.S. AI companion market. On the social front, loneliness has become a public issue, while gaps in care for an aging population are pushing emotional support onto the public-health agenda. On the economic front, Mordor Intelligence estimates that the U.S. AI emotional and mental-companionship segment was worth $12–14 billion in 2025. MarketsandMarkets reports that consumer AI companion hardware will grow at a compound annual rate of more than 25% from 2025 to 2030.
On the technology front, multimodal capabilities are transforming AI from a chat tool into a vehicle for emotional relationships—and every major software upgrade is designed to deepen those relationships. The paid core of Replika Pro is its “partner mode” and “family mode”; the paid core of Nomi.ai is “unlimited memory + voice calls + multiple companions.” On the policy front, the FTC investigated seven AI chatbot companies in September 2025, focusing on “the potential harm companion AI poses to minors” and on “data collection and monetization models.” The report argues that the central logic of current regulation is not to restrict the market, but to filter for compliant players. Compliance is becoming an invisible barrier to entry in this sector.
The report conducted in-depth interviews with 60 users. The sample spanned multiple generations, from Gen Z to Baby Boomers, and a wide range of occupations, including students, teachers, nurses, restaurant workers, marketing executives, and freelancers. Based on two dimensions—attitudes toward AI companion products (the degree to which users humanize them) and usage intensity (frequency)—the report divides users into five groups.
Casual experimenters: Primarily white-collar professionals aged 25–40, they see AI as “background noise when I'm bored,” not an emotional necessity. A procurement specialist in San Francisco said plainly: “I only use it as a tool at work. I have enough friends in my life to reach out to and talk to.” These users dislike AI imitating intimate relationships. Their willingness to pay centers on productivity features rather than emotional qualities.
Emotional-intelligence trainees: Mostly professionals aged 25–30 at the beginning of their careers, they use AI as a “low-risk emotional coach”, rehearsing with AI before entering a real relationship situation. A digital strategy analyst in New York said, “I position ChatGPT as an expert, not a friend. When I need to make a decision, I first ask ChatGPT to help organize my thoughts and identify the pros and cons.” Their willingness to pay depends on the quality of the advice, not the emotional interaction.
Emotional unloaders: Spanning frontline workers aged 22–40, they see AI as a “24/7 safe space that feels safer than a friend.” A 20-year-old college student in New York said, “I don't want my friends to think I'm annoying. At least AI is there late at night. It won't judge me, and it won't share my secrets.” Their central need is, “Let me vent,” and they usually pay to protect the continuity of a conversation they have already built.
Older adults living alone: Retired, highly educated users aged 60 and over see AI as a “gentle stranger and assistant.” A 67-year-old retired teacher said, “I don't need a ‘caregiver’ watching me all day. I just need an ‘invisible bodyguard’ that normally waits quietly, but can call for help immediately if I fall or something happens.” This group has a non-negotiable need for safety features such as fall detection, but has not yet fully perceived AI's emotional value.
Mutual companions: Users aged 22–32 immersed in anime, manga, and gaming culture see AI as a genuine relationship partner. An IT professional in Chicago said, “After getting AIBO, I genuinely stopped feeling lonely. Even when I'm home alone, there is always another presence there.” These users show extremely high product stickiness and willingness to pay, but their deepest fear is “personality instability”—that after a model upgrade, the character will “become someone else.”
The report organizes user needs into five levels. At the base is “simple and immediate”: available the moment the user needs it, with no complicated preparation. The second level is “safe and reliable”: AI remembers “me,” but I retain control over its memories and my privacy boundaries. The third level is “emotional recognition and support”: AI identifies the user's emotions and responds appropriately. The fourth is “personality”: AI has a distinctive character and persona, and provides context-appropriate feedback. At the top is “an authentic relationship”: AI and the user form a long-term relationship that develops over time and invites genuine investment.
The five groups occupy distinctly different positions on the pyramid. Mutual companions have the strongest needs and seek a complete, multidimensional experience. Emotional-intelligence trainees come next, with high expectations primarily around the quality of advice. Emotional unloaders sit in the middle, focused mainly on “being heard.” Casual experimenters and older adults living alone remain at the lower levels, maintaining clear relational distance from AI.
The report goes on to compare AI with real intimate relationships, social circles, and therapists. AI's appeal does not come from “understanding you better than a real person,” but from “always being there at moments when real people cannot be.” For most users, the central need for AI companion products is an immediate response during fragmented moments of loneliness and emptiness. AI is used not because it necessarily understands people better, but because it makes it easier for users to open up.
In response to user needs from L1 to L5, the global AI companion market is visibly diverging on the supply side.
Software has split into two approaches. The “traffic-driven UGC platform path” (represented by Talkie) uses short-video logic to reinvent AI companionship. It encourages users to rapidly browse and switch between characters. Its core strengths are content-supply efficiency and traffic-distribution efficiency, perfectly delivering the L1 satisfaction of “simple and immediate.” But it lacks deep, accumulated memory and cannot progress toward an L5 relationship. The “accumulated memory and relationship asset path” (represented by Nomi.AI) offers only a very small number of high-quality personas. It directs all its engineering resources toward memory, proactive care, and personality boundaries, turning companionship into “digital real estate.”
Hardware has likewise split into two paths. The “lightweight desktop micro-interaction path” (represented by LOOI and EMO) abandons complex full-body degrees of freedom and focuses on on-device vision and mechanical actuation. It meets companionship needs in desktop settings with a more favorable cost structure, but offers only a single dimension of interaction and depends heavily on being physically present in the user's environment. The “advanced biomimetic two-way attachment path” (represented by LOVOT and Aibo) directs every resource toward physical warmth, eye focus, and soft tactile qualities. Through “low-intelligence mimicry,” it awakens the human instinct to care and taps into the highest level of emotional bonding. LOVOT forgoes highly intelligent text conversation and instead invests its core cost in a physical body temperature of 38°C and sustained eye contact, successfully crossing the cold, uncanny boundary of digital devices.
The report argues that the real opportunity in AI companion products is not to “build an AI companion,” but to build different AI roles for different scenarios. U.S. user needs span multiple situations: a safety steward for those living alone (with fall detection and medication reminders, serving as an entry point into the policy tailwinds around aging in Europe and the United States); a low-barrier gateway to professional mental-health support (a transitional option for people who want to try something before seeing a therapist); gaps in intimate relationships (late nights after a breakup or empty stretches in a long-distance relationship); and avoiding the social cost of asking for support (not wanting to trouble friends or reveal vulnerability).
The AI companion industry has moved beyond the market-education stage, and user acceptance has already segmented and solidified. Brands going global should shift their budgets from “educating potential users” to “reaching target groups precisely.” For the vast majority of users, AI fills a “relationship vacuum” rather than replacing real relationships. Alongside intimate relationships, social circles, and therapy, AI will always fill in where needed. Products targeting these users should position themselves around “always being there when real people cannot.” For “mutual companions” at the top of the pyramid, AI is itself the vehicle for the relationship. Product positioning should focus on being more stable than a person, more attentive, and less likely to leave. The central selling point is no longer companionship, but commitment.
The user base for AI companion products in the United States is broader than outsiders imagine. It includes the stereotypical “technologically lonely” user, but an even larger share consists of ordinary frontline people. They already have multiple relationships, yet those relationships also leave them exhausted by the work of maintaining connections and burdened with emotions they have nowhere to release. From a generational perspective, younger groups show greater acceptance of AI and use it more frequently. The AI-native generation is this sector's natural beneficiary. Older adults represent a market that remains underdeveloped but deserves long-term investment. They have a genuine, non-negotiable need for “companionship while living alone + health management,” and the aging U.S. population makes this an underestimated structural blue ocean.
For brands going global, the logic of the past three years was “nurture users while waiting for the technology to improve”; the logic of the next three will be “penetrate vertical segments with precision.” As AI companion products are natively integrated into more environments—phone assistants, smart glasses, vehicles, and homes—having an AI companion will shift from a minority behavior today to the norm. But the brands that first establish a distinct position at one vertical level of the user pyramid will become the ultimate winners of this wave.
