“We’ve been sidelined.”
That’s how one artificial intelligence practitioner recently described the state of emotional companion apps to 36Kr. His team, part of China’s so-called “six tigers,” had developed an AI-driven emotional companion app that just a year ago was a regular top-three performer in its category and seen as a showcase for exporting Chinese AI abroad.
Today, that app has drastically scaled back. Most of its operations and monetization teams have been laid off, with only a small team left for basic upkeep. And it’s not the only one.
Emotional companion apps, designed to fulfill users’ emotional needs through interactions with AI-powered characters, once looked like a breakout category in generative AI. Known players include Talkie by MiniMax, ByteDance’s Maoxiang, and US-based Character.ai.
But the momentum has stalled.
In China, this space is showing signs of stagnation. According to Diandian data, Maoxiang’s monthly downloads on iOS fell from 2.64 million in January to 610,000 in May, while daily active users (DAUs) slipped from 590,000 to 490,000. Xingye, another major app, dropped from 4.86 million downloads to 930,000 during the same period.
The trend is mirrored overseas. Character.ai, widely credited with popularizing the category, boasts 233 million monthly active users (MAUs) but has struggled with monetization. Its average revenue per user (ARPU) is just USD 0.72.
At the beginning of 2024, emotional companion apps were one of the most sought-after investment themes. The concept had appeal: emotionally engaging AI, sticky user behavior, and subscription-driven revenue. Character.ai’s reported USD 15 million in annual recurring revenue (ARR) from a 12-person team became a much-discussed benchmark.
By mid-2024, nearly all of the “six tigers” had entered the field.
Now, nearly half of those products have already faded.
36Kr previously reported that Maopaoya, a prominent consumer-facing effort by StepFun, essentially shut down by December 2024 due to underwhelming performance. Zhipu AI’s AiU and Moonshot AI’s virtual pet app Ohai were both discontinued amid either regulatory hurdles or revenue challenges.
Beneath the early excitement, emotional companion apps have proven to be a low-return segment. AiU’s product manager, Yong Chao, stated plainly: “It might look lively, but not many products have truly delivered strong results. In China, there are only five apps with over 10,000 DAUs.”
According to 36Kr, one leading app in the category brings in under USD 40,000 in daily global revenue, a far cry from the publicly floated eight-figure projections.
That modest income is often overshadowed by massive marketing costs. One such product reportedly burns through an eight-figure USD marketing budget each quarter.
The poor return on investment is partly due to model limitations, such as weak long-term memory and inconsistent narratives, that hinder deeper emotional engagement. As a result, user stickiness and willingness to pay remain low.
A QuestMobile report found that in the first half of 2024, users interacted with mainstream emotional companion apps fewer than five days per month on average. One investor questioned the basic premise: “At the end of the day, people still prefer real relationships.” If emotional companionship is indeed a “false need,” then the market for it is inherently limited.
Per 36Kr, one of the top apps in the category generates only USD 400,000–500,000 in monthly user payments. Attempts to boost revenue by introducing additional paid features have had limited effect.
Meanwhile, high-pressure growth targets and intense competition pushed companies to unsustainable tactics in 2024. Some set aggressive goals like reaching one million DAUs, and spent heavily across dozens of ad channels to get there.
The results didn’t hold. “At one point, second-day retention hit 50% thanks to heavy marketing, but it quickly fell back to 20% or 30%,” one employee said. “It has never once been positive” in terms of ROI.
Then there’s regulation, the constant threat hanging over the category.
It has already led to the removal of high-profile apps like Talkie, LiveMe, and Tango from app stores globally. In June, Zhumengdao, a product from China Literature, was taken down by regulators for suggestive content involving minors.
These takedowns halt growth abruptly. Multiple sources told 36Kr that one app’s organic growth rate dropped from 20% to under 10% after being reinstated, with its MAU declining by about 10% each month.
To meet compliance demands, many teams resorted to sweeping word filters that compromised usability. A staffer at one of the major emotional companion apps cited an example: to avoid regulatory trouble, its team added “take off” in Mandarin to the list of banned words. As a result, even associated phrases like “take off a hat” would trigger conversation blocks or account suspensions.
In 2025, as the hype fades and realism sets in, those still competing in this space are entering a phase of reflection and regrouping.
Sources told 36Kr that the leading app mentioned earlier has cut its marketing budget by nearly 90%. “Now we focus on optimizing for returns. We invest more in channels or periods where the returns are actually good,” an employee said.
With technical constraints still in place, it’s also time to reconsider the core proposition of these apps. One investor noted that both emotional support and virtual friendship may be concepts with limited long-term appeal.
Character.ai, for instance, has repositioned itself as a content-driven community. Since its acquisition by Google in August 2024, it has focused on multimodal tools for image and video generation rather than just chat-based interaction.
Still, not all signs are bleak.
HiWaifu, an emotional companion app that earned more than RMB 20 million (USD 2.8 million) in 2024, offers a glimpse of what might work. Founder Tao Jiang found that apps using realistic human images monetize better than those with anime-style avatars. “We’re now exploring directions like emotional support for seniors,” he said.
Thanks to modest marketing spend and a lean team, HiWaifu reportedly became profitable within its first year.
As one investor summed it up: Character.ai still has over 200 million MAUs. With a user base that large, monetization is not the core challenge.
“The truth is, no one has really nailed the product or the business model for emotional companionship,” the investor said. “This space isn’t dead, people just haven’t unlocked its full potential yet.”
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Zhou Xinyu for 36Kr.