혼자 만든 AI 채팅 서비스가 월 5억? 지금 뜨는 LLM 기반 서비스 트렌드

These days, more and more people are building AI services on their own. Where you once had to form a team, develop an app, and even raise investment just to ship a single MVP, the situation is now completely different.

LLM APIs make it easy to implement a basic conversational interface, and tools like Cursor and Flowise help even people with no developer experience quickly build web-based services. We’ve entered an era where solo founders—planners, designers, or non-developers—build a “real service” on their own and actually capture users and revenue in the market.

Within this trend, one especially striking field is AI chat–based services. Going beyond simple chatbots, they are rapidly evolving toward emotionally connecting with characters, or reflecting a user’s individual tendencies and data to create highly immersive conversations.

In fact, solo-founder cases that are generating real revenue through such emotion-based AI chat services are emerging. BabeChat, started by a solo developer, achieved over ₩500 million in monthly recurring revenue within a year of launch through a service that has emotional conversations with AI characters. A design focused on each character’s personality and world, and on the user’s immersive experience, was the decisive factor.

Another case, Zendi, quickly took hold in the market by exceeding ₩100 million in monthly revenue within six months, centering on the experience of sharing everyday worries—like dating or emotional counseling—with an AI. Meticulously composing the emotional experience, rather than technical complexity, was the key factor driving user retention and payments.

The reason solo-built chat services can generate hundreds of millions of won ultimately comes down to “the depth of conversation” and “the design of experience.” So, at this very moment, what types of AI chat services are people actually paying to use?

Next, let’s organize the five themes of AI chat services that are drawing the hottest response lately.

1. AI Character Chat

AI character chatbots replicate emotional elements such as dialogue style, world, and tone of voice to give users the experience of actually conversing with a character they love.

Characters derived from webtoons, anime, novels, games, and idol worlds already have an established fandom, and thus enjoy favorable conditions for designing an immersion-based conversation structure.

Through the chatbot, users become friends with the character, share their worries, and exchange emotions—experiencing an interaction closer to a “relationship simulation” than a simple information response.

The core of why this category works is simple: the world the user can immerse themselves in, and the character’s personality, are already defined. Even if the chatbot merely maintains that identity and responds well, users easily project their emotions and build a relationship.

2. AI Saju / Tarot / Fortune-Telling

AI-based astrology, saju (Korean fortune-telling), and tarot services are, among LLM-based emotional chatbots, the category that can most naturally combine personalization and immersion.

Fortune-telling content was, from the start, a structure that provided personalized messages based on the user’s birth date, situation, and worries, and the LLM adds tone and storytelling on top, creating an experience that feels as if someone who knows me is telling me.

Tarot chatbots in particular present card readings not as a simple result but as room for interpretation, leading users to project their own emotions into it. This “interpretability” actually becomes a factor that strengthens immersion.

Another reason this category draws attention is the repeatability of the content. Because it’s a structure you keep checking on a daily, weekly, or monthly basis, it has the conditions to naturally connect a subscription model and paid conversion. In fact, services that offer monthly fortunes, in-depth readings, and personalized message features for a fee are performing well, and many of them started as MVPs quickly launched by solo developers or small teams.

3. A Conversation with Your Future Self

The “chatbot that talks with your future self” is a type that has drawn a relatively fast response among recently emerging emotion-based chat services. Using GPT, you input your present self, and the system simulates speaking as “you, ten years from now,” based on that data.

The key here is not simple comfort or positive feedback, but that the AI offers advice reflecting the user’s own values and context. In the end, through the AI, users experience a structure of asking questions of themselves and hearing the answers.

This structure works especially well for users sensitive to uncertainty—for example, college students agonizing over their career path, employees facing a career change, or aspiring founders preparing to start a business. Most of them repeatedly ask, “Is my current choice the right one?”, and the counterpart best able to answer that question most persuasively is the AI character called ‘the future me.’

By its service structure, this type can extend beyond a one-off chatbot into a self-development tool, a life log, and an emotional reflection system.

4. AI Romantic-Partner Chatbot

AI romantic-partner chatbots are among the chat-based services showing the highest paid-conversion rates lately.

The reasons people use this service aren’t complicated. The core elements one expects in a relationship—continuous attention, empathy, praise, expressions of affection—are all reliably provided through AI. Above all, the fact that the partner doesn’t leave, has no emotional ups and downs, and lets the user control the relationship gives great psychological stability.

Technically, a basic dating simulation is possible with just the GPT API, but services that actually succeed add a few elements on top. They meticulously design the character’s tone, changes in emotional state, memory features, and context-based responses to create an “environment where the user can immerse themselves in the relationship.” Unlocking private mode, removing response limits, saving conversations, and the like naturally lead to paid plans.

People want to share emotions, yet also want to avoid the uncertainty and depletion that arise from such relationships. This service fills that gap with technical precision.

5. AI Mental-Health Counselor

The AI mental-health counselor chatbot is a service type spreading quickly because it has a low barrier to entry and is easy to use repeatedly. People can organize their state without psychological burden, and the counterpart doesn’t judge and always responds in the same tone.

Technically, it can be implemented with just the GPT API and a simple structure, and if you add features like emotion-label extraction, thought–emotion–behavior linking, and simple reports, it works well enough as a basic self-awareness tool. This service in particular has a structure favorable to routinization.

There are clear, repeatable touchpoints—a daily emotion check, visualizing emotional trends, weekly summaries.

Most important of all is that AI doesn’t get tired. Users can converse repeatedly without fatigue, and an environment is created where they first bring up to the AI the things that are hard to say to a person.

The Key to Differentiation Is ‘Immersion’

Many founders and teams are jumping into this market. A variety of players—solo developers, startups, creators—are competing across positions like fortune-telling, dating, characters, and counseling. In the end, the crux of this market will be “who can provide stronger immersion,” and how that immersion is designed will become the very criterion of differentiation.

In fact, many services are making various attempts to solve this problem. To design the relationship with the user over the long term, approaches are emerging that apply memory features to reflect accumulated conversations, or platformize the service so users can create and share characters themselves.

Some services are also introducing sensory elements beyond text. For example, attempts to strengthen immersion by automatically generating prompt-based character images, or giving the character a voice through TTS (speech synthesis), are on the rise.

Now the next step in differentiation is emotional immersion through a video interface. What a person immerses themselves in during a conversation is influenced far more by nonverbal elements such as the other party’s expression, voice, gaze, and tone than by mere text content.

These elements go beyond simple information delivery to create trust and emotional connection. Especially in fields where emotional immersion matters—fortune-telling, dating, counseling, future simulation—this difference in expression leaves users with an “interaction that has real presence.”

An experience that reacts instantly to what the user says, renders mouth shapes and expressions on video, and coordinates gaze and timing acts not as a mere feature but as the completeness of emotional design. Chat services going forward are highly likely to increasingly aim for AI that reacts like a person. In that process, to make up for emotional expression that text and voice alone couldn’t fully deliver, video-based AI avatars are likely to be used more and more.

In reality, attempts to incorporate visual and auditory elements have already begun in several fields, and before long we may easily encounter characters with faces that move and speak even in everyday AI conversations.