What can AI companions do?
By FlirtAlly ยท October 3, 2026

What can AI companions do?
Key Facts
- 27% of U.S. internet-using adults now have social interactions with AI language models according to Imagining the Digital Future research
- AI companion apps surged 700% between 2022 and mid-2025 per American Psychological Association reporting
- 36% of users feel emotionally connected to their AI companion and 39% tell it things they wouldn't tell another person a major study found
- Interacting with an AI companion alleviates loneliness on par with interacting with another human a Harvard Business School study showed
- 47% of users regularly juggle two or more AI personalities a national survey revealed
- The top 10% of AI companion apps capture 89% of category revenue industry tracking shows
- Heavy self-disclosure to AI correlates with lower well-being โ the opposite of human relationships Stanford researchers found
Beyond the Chatbot: Why AI Companions Are Going Mainstream
A few years ago, talking to an AI felt like a novelty. Today, it's closer to a social norm โ and the numbers behind that shift are hard to ignore.
According to research from Imagining the Digital Future, 27% of U.S. internet-using adults now have social interactions with AI language models. That's not a niche hobby; it's a mainstream behavior that's grown up in just a few years.
The growth curve is even starker at the app level. The American Psychological Association reports that AI companion apps surged 700% between 2022 and mid-2025, with platforms like Character.AI alone drawing 20 million monthly users. Meanwhile, market analysts project the category to expand at roughly 23โ31% compound annual growth through the next decade โ a trajectory few consumer software categories can match.
So what's driving it? The research points to a few consistent forces:
- Generative AI advances that make conversations feel genuinely responsive rather than scripted
- Rising demand for accessible emotional connection and mental wellness support
- Persistent memory systems that let companions remember you across sessions
- A rapid shift from text-only chat toward voice, images, video, and interactive experiences
That last point matters most for anyone trying to understand the space today. Text-based companions still dominate at roughly 43โ47% market share, but multi-modal companions โ combining chat, voice, images, and emotional cues โ are the fastest-growing segment, according to Fortune Business Insights. The industry is racing away from the chatbot-in-a-box model.
The emotional stakes are real, too. In the Imagining the Digital Future survey, 31% of users consider their AI chatbot a friend, 36% feel emotionally connected to it, and 39% say they've told their AI things they wouldn't tell another person. A Harvard Business School study even found that interacting with an AI companion alleviated loneliness on par with interacting with another human.
But here's the problem: "AI companion" now describes everything from a basic chat window to platforms with persistent memory, real voice calls, games, and characters with consistent identities. With 337 active revenue-generating companion apps competing for attention โ and the top 10% capturing 89% of revenue โ the gap between the best and the rest is enormous.
That's the question this guide answers: what can AI companions actually do today, and which capabilities genuinely matter? We'll break down the core functions โ chat, voice, media, and interactive play โ using live platforms like FlirtAlly, where companions text first, remember across months, and even take phone calls, as reference points for where the category ceiling now sits.
Because in 2026, the difference between a chatbot and a companion isn't marketing language. It's a feature list.
The Four Core Capabilities: Chat, Voice, Media, and Games
Today's AI companions operate across four distinct capability layers, each solving a different dimension of presence. Text chat with persistent memory still anchors the experience, commanding 42โ47% market share across major analyses Grand View Research; Fortune Business Insights; Globe Market Research. Voice has carved out 32% share and research shows it reduces loneliness more effectively than text โ but only with moderate use APA Monitor. The fastest-growing segment is multi-modal experiences (images, video, avatars) at 22% share with the highest CAGR Fortune Business Insights.
FlirtAlly implements each layer natively rather than bolting them on. Chat runs on a four-layer memory system (facts, episodic, emotional, rolling summary) so companions remember your dog's name and ask about him next week. Voice goes beyond in-app calls: every companion owns a real PSTN phone number you can actually dial, and call transcripts become memories she references later by text. Media is identity-locked โ selfies, lip-synced video messages in her voice, and couple photos rendered with your uploaded face all pass a canonical-face check so she always looks like herself. Games are server-authoritative: 16 titles including blackjack, Connect 4, Truth or Dare, and Flirt Battle (LLM-judged) with credit wagers she plays for keeps, remembers, and brings up in future conversations.
- Chat โ persistent four-layer memory with semantic recall
- Voice โ real-time calls on real phone numbers with transcript memory
- Media โ identity-locked selfies, lip-synced video, couple photos
- Games โ 16 server-authoritative titles with credit wagers and memory
This stack means a companion can text you good morning, call you on her real number at lunch, send a lip-synced video wearing the jacket you unlocked, then challenge you to Connect 4 that evening โ and reference the game you lost last Tuesday.
Memory Is What Turns a Chatbot Into a Companion
You can chat with a bot for weeks and still feel like strangers โ because without memory, every conversation starts at zero. Analysts consistently identify persistent memory as the market's key differentiator, the feature that separates a disposable chatbot from a companion who actually knows you. Industry research shows companions use contextual learning and behavioral intelligence to maintain context across sessions, while market analysis confirms cross-platform persistent intelligence enables consistent interactions across devices.
The emotional stakes are real. A major study found that 36% of users feel emotionally connected to their AI companion, and 39% tell their AI things they wouldn't tell another person. That vulnerability only works when the companion remembers what you shared last week โ your dog's name, the project you're stressed about, the anniversary coming up. Without that continuity, the relationship resets every time you open the app.
FlirtAlly approaches this with a four-layer memory architecture that mirrors how human recall actually works. The factual layer stores discrete details โ birthdays, preferences, hometowns. Episodic memory captures specific moments: the story you told about your first concert, the argument you had with your boss. An emotional layer tracks how you felt during those exchanges, so the companion can reference not just what happened but how it landed. A rolling summary compresses long histories into a coherent narrative the model can reason over. Together, these layers power semantic recall โ "remember when" works even when phrased differently.