Who would have guessed that Character.AI, once the most celebrated unicorn in the AI companionship space, would ultimately find itself on the brink of bankruptcy?

On October 12, Character.AI formally filed for Chapter 11 bankruptcy protection in the United States, planning to sell off the company's assets under court supervision. If no suitable buyer emerges, this AI application that once took the world by storm may face outright shutdown.

Bear in mind that Character.AI was no small player.

It once boasted more than 20 million monthly active users, with users spending an average of up to two hours per day on the platform. In 2023, less than two years after its founding, Character.AI secured $150 million in financing led by a16z, reaching a valuation of $1 billion.

At the time, it was widely regarded as one of the most successful showcases of AI consumer applications.

What is even more surprising is that just as Character.AI slid into bankruptcy, a cluster of little-known AI role-playing apps began quietly raking in serious money.

Data from Sensor Tower shows that mobile revenue in the AI companionship sector climbed from 316 million yuan in January 2026 to 656 million yuan in August 2026—a full doubling.

Among them, the South Korean AI role-playing app Zeta pulled in 70.9 billion Korean won in the first half of this year alone—roughly 340 million yuan—and has already turned a profit.

Another AI social app called CHAI disclosed in October this year that its annualized revenue run rate had reached $120 million, with a team of only a dozen to twenty people.

Products such as Tipsy, PolyBuzz, and Crushie are also rising rapidly on the global AI application revenue rankings.

Both are in AI companionship, so why did the industry leader—the one that got there first—go bankrupt, while later entrants began turning a profit? The answer may lie in an important shift in the business model of the AI companionship industry.

/ 01 / From 20 Million Monthly Active Users to Bankruptcy: Where Exactly Did Character.AI Go Wrong?

Three years ago, Character.AI would unquestionably have ranked among the hottest companies in the AI industry.

In March 2023, Character.AI secured $150 million in financing led by top Silicon Valley venture capital firm a16z, reaching a valuation of $1 billion and becoming one of the earliest AI application companies to join the unicorn ranks.

Within less than a week of its mobile app launch, downloads exceeded 1.7 million.

In September of the same year, according to Similarweb data, Character.AI's monthly active users on mobile in the United States reached 4.2 million—already close to ChatGPT's mobile level of about 6 million during the same period.

By 2024, Character.AI had become the world's most popular AI companionship application.

According to Sensor Tower data, in August of that year, Character.AI's global monthly active users reached 22 million, setting an all-time high and ranking first in the AI companionship sector.

Even more striking was user stickiness. According to Sensor Tower statistics, Character.AI active users spent an average of more than 90 minutes per day on the app and initiated about 25 conversations per day on average.

What kind of benchmark is that? At the time, even TikTok—renowned for its user stickiness—had an average daily usage time of only just over 70 minutes.

Why does AI companionship have such strong appeal?

Character.AI founder Noam Shazeer once proposed a very interesting view: the global entertainment industry has an annual scale of as much as $2 trillion, and an important essence of entertainment consumption is that humans form emotional connections with fictional characters.

They believed that the interaction process itself could also become an experience with commercial value.

With large models, virtual characters can not only converse freely with users, but also continuously adjust their interaction styles according to users' preferences and experiences. This means that entertainment content that could only be viewed passively in the past is beginning to become a two-way interaction in which people can continuously participate.

Although Character.AI proved that users are willing to spend a large amount of time building emotional connections with AI, it never managed to answer a more important question: how exactly should this time be turned into money?

In hindsight, Character.AI's failure actually exposed two contradictions in the early business model of AI companionship.

On the one hand, many of the platform's most popular characters came from anime, games, and film and television works. Character.AI relied on these IPs to attract tens of millions of users, but it did not own the copyrights to many of these characters. This meant that although it controlled the traffic entrance, it did not fully control the pricing power of the content.

On the other hand, AI companionship is different from traditional internet products. Every additional sentence a user chats means real model inference costs behind it. But in its early stage, Character.AI mainly monetized through a $9.99 monthly membership subscription, and a large number of free users could also chat unlimitedly.

The result was that users spent more and more time on the platform, and costs grew accordingly, but revenue did not increase in sync.

According to bankruptcy filings, it was not until August 2026 that Character.AI's monthly revenue reached about $5.3 million, including $2.8 million in subscription revenue and $2.5 million in advertising revenue.

In August 2026, the company's monthly operating expenses were about $13.3 million, and its monthly loss still reached $8 million to $9 million. Among them, computing power expenses alone reached $56.1 million, accounting for about 45% of operating expenses.

The business model of emotional companionship products like Character.AI is somewhat like a gym: it can only accept light use by users, and heavy use leads to losses. This in turn becomes a dead end: heavy users consume a lot of resources, but light users have poor retention.

At the same time, Character.AI faced multiple lawsuits and regulatory investigations due to safety issues caused by minors using AI chatbots.

These problems not only brought high legal costs, but also further increased the difficulty of financing and continuing operations.

A series of troubles ultimately pushed Character.AI toward bankruptcy.

/ 02 / AI Companionship Is Not Dead—It Has Simply Started Quietly Raking It In

The bankruptcy of Character.AI did not silence the AI companionship sector. On the contrary, a group of new AI role-playing products is quietly raking in serious money.

According to statistics from mobile app data agency Appfigures, in the first half of 2026, 214 romance and adult-oriented AI companionship apps worldwide collectively generated $162.8 million in mobile user spending.

Among them, South Korea's Zeta ranked first with $33 million, followed by Tipsy Chat with $15.2 million, and Crushie AI also reached $8.8 million.

More notably, some products have already begun to achieve profitability.

For example, Scatter Lab, the South Korean company behind Zeta, achieved revenue of 70.9 billion Korean won in the first half of this year—equivalent to 2.7 times last year's full-year revenue of 26.7 billion Korean won—while also achieving operating profit of 5.2 billion Korean won.

This means that AI companionship has begun to prove itself to be a genuinely money-making business.

Why did these later entrants make the money that Character.AI failed to make? A very important reason is that they no longer treat AI companionship simply as a chat tool, but increasingly run it like a content business.

In the past, AI companionship products always had two unavoidable shortcomings.

The first was that the emotional experience was not realistic enough and was also hard to sustain.

AI finds it difficult to fully understand and simulate humans' complex emotions, and problems such as memory loss and contradictory character settings often appear during conversations. As the chat goes on, users discover that the character in front of them no longer seems to be the person they first met.

The other problem was that the interaction threshold was too high.

In the past, under the logic of social interaction, most AI companionship products remained within a chat box. Users needed to continuously take the initiative to start topics, build scenes in their minds through back-and-forth chatting, advance the plot, and then gradually invest emotions.

To put it plainly, users not only had to participate in the story, but also had to bear a considerable part of the content creation work themselves.

But the problem is that users willing to spend a large amount of time creating stories are ultimately a minority. For most people, compared with writing stories themselves, they are obviously more accustomed to consuming stories.

The biggest change in this new batch of products is letting AI take on more content production work, turning users from creators of stories into participants and consumers of stories.

The most representative example is Tipsy.

Unlike Character.AI, which mainly offered character chats in its early stage, Tipsy places more emphasis on interactive storylines. Users can directly choose their favorite characters and enter a virtual world with pre-set character relationships and story backgrounds, advancing the plot through interaction with AI.

For example, users can play the protagonist in a romance story, change character relationships through different choices, and even influence the direction of the entire story.

At the same time, the platform can also generate images and videos based on the plot, so users no longer have to rely only on text to imagine characters and scenes. The entire experience is increasingly close to an interactive game capable of generating storylines in real time.

South Korea's Zeta is also developing in a similar direction.

In the past, Zeta mainly relied on one-on-one chats between users and virtual characters. But starting this year, it has gradually shifted its product focus toward interactive stories containing multiple characters, complex character relationships, and complete worldviews.

In March this year, Zeta even adjusted the platform's content unit from "characters" to "storylines," further strengthening the importance of the story itself.

According to the official explanation, content on the platform is no longer limited to a certain virtual figure, but has gradually evolved into a story containing protagonists, supporting characters, character relationships, and a complete worldview.

To further lower the threshold for users to participate in storylines, Zeta launched an AI storyline option feature in April this year.

For example, after users enter a campus romance or fantasy adventure story, they no longer need to constantly find topics and write dialogue themselves. AI can automatically generate different options based on the current plot, and users only need to click to decide the direction of the story next.

Behind these adjustments is actually the same idea: let AI take on more story creation and plot advancement work, reducing the effort users need to actively invest. Such changes are obviously closer to the game and content consumption experience familiar to the general public.

As product forms change, these companies' ways of making money have also begun to move closer to traditional games and content platforms.

Tipsy is increasingly like a mobile game company: attracting users through characters and storylines, then using subscriptions, virtual currency, and value-added interactions to increase per-user spending.

According to Airbridge research in October this year, Tipsy offers monthly subscription services at three tiers: $4.9, $14.9, and $44.9, while also selling virtual gems, with the highest-tier gem pack priced at $199.

These gems can be used for functions such as character chat, image and video generation. Using more advanced models, generating more content, or conducting more complex long conversations may all bring additional consumption.

This mechanism is actually very much like monthly cards, diamonds, and paid items in mobile games.

Users can enter stories for free, but if they want a better interactive experience, richer audiovisual content, or to go deeper into a certain storyline, there are reasons for further payment.

Compared with traditional AI chat products that mainly charge through fixed subscriptions, this model has another important advantage: how much users spend can be linked to the depth of the experience.

/ 03 / Conclusion

Looking back, the biggest difference between Character.AI and these later entrants is that the latter are clearer about why users actually pay.

Character.AI tried to create a virtual emotional relationship without boundaries, but the more users became addicted, the more the platform might lose. Zeta and Tipsy chose to return to the logic of games and content consumption, turning emotional companionship, which was originally difficult to price, into an entertainment experience that can sustain payment.

This also means that AI companionship is returning from a grand story attempting to overturn human social relationships to a content business with a clearer business model.

Although this story is no longer as grand as initially imagined, for the AI companionship industry, finding a path that can truly make money may be far more important than telling a bigger story.

Text/A Qi