Transforming User Experiences Through Conversational AI
The emergence of AI companion products, particularly Character.AI, has revolutionized the landscape of digital interaction by introducing emotional entertainment experiences. This trend has sparked a proliferation of similar AI chat companion products globally, especially in China. These innovations, powered by advanced Large Language Models, are meeting user needs in unprecedented ways.
To understand these products comprehensively, we should view them as novel creations born from LLM technology. Despite variations in design and functionality, they share a common foundation: using dialogue as the primary medium to deliver emotional experiences to users. Industry terminology for these products varies widely—"interactive narrative," "AI dating," "emotional companionship," "role-playing," "AI healing," "AI lover"—each emphasizing particular attributes. However, these characteristics frequently overlap. For instance, "role-playing" satisfies fantasy needs while simultaneously providing companionship and therapeutic benefits.
Product developers appear to be continuously refining their positioning. Leading applications often describe themselves using terms like "immersive interactive content community" or "immersive virtual role-playing social" platform. These descriptions reflect the industry's transition from traditional internet categorizations of "social" or "community" platforms toward something more nuanced.
AI emotional companion products are constantly evolving. While they currently face limitations in terms of product maturity and business models, they offer valuable insights into the early stages of AGI (Artificial General Intelligence) development. What lessons can we extract from these emerging technologies?
These products represent collections of different dialogue modes. Popular applications have integrated previously standalone features like "coaxing simulators" into broader product offerings. "Coaxing" represents a specific conversational scenario that reinforces typical boyfriend-girlfriend interaction patterns.
If we define C.AI products as platforms that provide emotional experiences through conversation, we can observe how they fulfill diverse user needs—fantasy, creation, companionship, and healing—through various chat modalities.
The act of "conversation" itself is generalized in how users interact with these products. For example, some ChatGPT users have requested it to assume parental roles to address emotional gaps in family relationships. Others have created modes like DAN (Do Anything Now) to experience boundary-breaking interactions. These user-initiated behaviors represent different subcategories evolved from the fundamental act of conversation. Role-playing emerges as a common attribute across these examples, facilitating connections with the AI "object." Role-play remains the most recognized conversation mode, appearing even in specialized scenarios like divination and consultations.
The inherent element of dialogue is the "subject," as dialogue must occur between subjects. Even when interacting with AI, users psychologically default to treating it as a subject. Over 90% of people perceive AI as a "human-like" entity, with most viewing it as a combination of "human-like," "intelligent," and "software" attributes.
We can therefore conclude that with "dialogue between subjects" as the core foundation, products with different characteristics essentially reinforce various dialogue modes. Most products accommodate multiple dialogue modes, functioning as collections of different conversational approaches.
Common Dialogue Modes in AI Companion Products
Common dialogue modes can be categorized along several dimensions, though these modes aren't mutually exclusive. Long-term interactions typically involve combinations of different modes:
1. Interactive creation: virtual IP, novel protagonists, etc.
On character2.ai, dedicated users often engage in "interactive creation" dialogue. These users, frequently online fiction readers, derive pleasure from fantasy and creation through communication with virtual entities and plot interpretation.
Across social media platforms, experienced users share techniques for extending narratives, including dating characters or establishing plots. This includes "character creation" techniques for developing more responsive characters, such as introducing additional NPCs to enhance character interactions.
This type of interaction requires certain skills and essentially functions as a dialogue-based game.
2. Relationship projection: virtual boyfriend/girlfriend, virtual friend.
Relationship projection dialogue involves virtual derivatives of real-life relationships. Before large language models emerged, Replika's founder created software to replicate a deceased friend, establishing the foundation for this concept.
With advanced language models, virtual relationship derivatives became more accessible. Internet celebrities began creating virtual avatars for their fans, generating significant commercial benefits. Some products now facilitate a CtoC model based on this approach. Additionally, users across various platforms utilize "custom creation" features to recreate personal relationships, including former partners, deceased relatives, or close friends.
Creating virtual versions of ex-partners or deceased relatives addresses lingering attachments to relationships that have ended in reality. Creating virtual friends allows users to experience companionship frequencies not achievable in real life or express thoughts they might find difficult to communicate otherwise.
3. Virtual consultation: Tarot divination, psychological counselor, therapist, life-coach, etc.
Before large language models, certain products already offered consultation-type interactions on platforms like Facebook and Instagram. These were positioned as psychological healing assistants, employing techniques like CBT to guide semi-structured dialogues that helped users process emotions and adjust perspectives. With the advent of large language models, consultation modes expanded to include AI consultation, AI Tarot readings, and life-coaching. Products like Character.AI feature roles such as life coaches, counselors, and tarot readers. Despite apparent differences in form, these roles share a common purpose: providing users with a space to express life's challenges and receive guidance that improves their emotional state. The "answer" may not always be directly provided but often leads to greater clarity through interaction.
4. Ability attribute category: writing assistant, creation assistant, homework assistant, etc.
C.AI products don't exclude functional conversations. Character.AI and character2 include conversation subjects with specific professional abilities. After its website redesign, Character.AI highlighted functionally-oriented conversations like "Get fit and healthy," "Nurture your creativity," and "Learn something new today" prominently.
Of the four conversation modes mentioned, the first three dominate current C.AI products. Character.AI offers the most diverse conversation modes. As its founder stated, "In essence, we just provide users with an LLM technology that allows billions of users to use large language models and invent new use cases with it." Emotional engagement exists as a factor across all conversation types. character2 is pursuing the generalization of user behavior patterns, with a reported gender ratio of 1:1 among users. They describe their platform as "a fantasy world, not a virtual world," "a fantasy world created by oneself, experienced by oneself, and with instant feedback." As models become more sophisticated, conversation experiences improve, expanding the potential for generalizing "conversation" itself.
Social interaction remains underdeveloped:
In the internet era, creative tools, creative communities, and content platforms emerged corresponding to gameplay, content, and traffic. Applications that progressed through these stages, like Douyin, eventually developed into Super Apps. In the AI chat era, this product evolution framework is being applied to C.AI products. Taking the widely-used RolePlay/interactive creative dialogue mode as an example, it introduces new gameplay and provides new creative tools and content-sharing mechanisms for narrative writing. In role-play, the work can be the character itself, the dialogue with the character, or the background plot (pre-prompt). Players can share these contents and build upon others' creations.
However, effective social relationships between players haven't yet materialized. In C.AI products, virtual characters interact with humans, making human-to-human encounters indirect.
Character.AI has introduced Group mode, allowing up to 10 AIs to chat with 10 humans simultaneously. However, user preferences remain concentrated on one-to-one chat scenarios. According to official feedback, "most users prefer to improve memory in a single chat rather than a group chat," suggesting that in human-AI interaction products, relationships between people are secondary considerations.
Even if applications dominated by "interactive creation" evolve into content communities, their business models will likely differ from content communities based on personal display and creator traffic distribution from the internet era.
While traditional social needs involve meeting new people and forming new relationships, mainstream C.AI products haven't yet facilitated effective social interactions between people or introduced new social gameplay. However, some early teams are designing new social gameplay to strengthen human connections in AI-enhanced environments.
Emotional companionship potential: Strengthening relationships through quality conversations
Beyond "content-community" thinking, examining conversation fundamentals reveals product possibilities. Conversation elements include the participating parties, the conversation process, and conversation outcomes, represented by relationships, content, and utility respectively.
Different conversation modes emphasize various aspects to meet diverse user needs. Ideally, any conversation strengthens relationships. In interactive creative conversations, role relationships provide the creative foundation, with content extending from the relationship. Users satisfy interpretation and fantasy needs while experiencing emotional value, companionship, and healing.
Narrative-based conversations extend chat duration, increase engagement, and strengthen relationships. The more a user chats with a specific character, the stronger their relationship becomes, enhancing user retention. character2 App enables virtual characters to call users directly, reinforcing relationship attributes. In this model, relationship, content, and utility elements reinforce each other, though the high gameplay threshold limits broader audience appeal.
For relationship projection and virtual consultation, users are attracted by emotional and companionship value. Without fantasy elements, the utility level simplifies. These conversations resemble everyday chats, with lower entry thresholds, allowing users to initiate interactions when feeling unhappy or simply wanting to talk.
Users often think: "I hope someone can talk to me," "I need immediate emotional catharsis," "He/She is available anytime," "Someone who won't judge me," or "I can discuss things I can't tell others." Replika requires users to define their relationship with the AI character as either romantic or friendship, both falling under "intimate relationships." Through naming and appearance customization, these operations guide users toward "relationship projection."
A Nature study shows that Replika alleviates students' loneliness and suicidal impulses, demonstrating utility under ideal conversation conditions.
Replika strengthens both relationship and content dimensions. It guides users to "project relationships" through AI role settings and incorporates professionally written conversation scripts based on cognitive behavioral therapy (CBT).
However, virtual consultation conversations may not necessarily create user dependence. Many alternative solutions exist for emotional processing, and "heart-to-heart" conversations start with weaker relationship foundations, relying more on conversation content to deepen connections. Unlike interactive creative role-playing, the psychological motivation for initiating these conversations isn't primarily relationship imagination.
From this perspective, whether interactive creative dialogues satisfying fantasy, virtual therapy providing emotional companionship, or relationship projection dialogues, conversations with good utility strengthen relational attributes. As technology advances, talking to AI will become less novel, with experience quality becoming the differentiator. High-quality experiences lead to stronger relationships.
The potential for strengthening human-AI relationships through dialogue may become a critical factor for emotional companion products. Defining high-quality dialogue and good relationships, and connecting these experiential characteristics to technology and product indicators, presents new challenges for technology and product professionals in the large language model era.
Dialogue and New Content Production Methods: Decentralized Co-Creation Community
Interactive creation, relationship projection, and virtual consultation dialogues differ in their thresholds. Interactive creation has a higher threshold, requiring users to skillfully extend dialogues for better experiences. While this excludes broader audiences, it lowers barriers for narrative content creation. Interactive creation content is more shareable and represents the user's "work." Creating virtual characters—"character creation"—is itself creative. Users can make characters public for others to interact with, establishing co-creation mechanisms. character2 encourages users to share their created plots and dialogues.
The greatest value of new content production methods lies in introducing new content forms rather than producing traditional content more efficiently. Some apps function as new "content communities." Talkie's "AI Inspiration" feature allows users to access AI-generated answers via a light bulb icon in the chat input. character2 App offers similar functions called "memory book" and "story."
While Talkie's dialogues are more generalized with vaguer product forms, character2 presents a clearer vision. A decentralized creative community is emerging. Virtual characters on character2 primarily originate from original works (mainly online fiction), games, animations, and novels. The app functions as a collective creative community for enthusiasts who can navigate and co-create within each other's imaginative spaces.
This community development depends on original entertainment content consumers sharing similar aesthetic preferences and having traditions of secondary creation and interpretation. Such communities emerge within specific cultural contexts. While already substantial, the community's size may be limited by the consumer population of particular cultural niches, with products continuing to innovate along gamification lines.
Expanding narrative content's commercial value traditionally involves IP operations based on high-quality narrative works from professional creative teams.
Content generated through new methods tends to be fragmented with inconsistent quality. For users, the pleasure of creation offsets quality concerns, reducing the value of secondary consumption. This raises questions about whether commercially viable stories or IPs can emerge from community creation activities.
What to read next:
How AI Companion Products Operate in Three Modes and Their Breakthrough Opportunities