Things like Character AI β what changed and where users went
By FlirtAlly Β· October 7, 2026

Things like Character AI β what changed and where users went
Key Facts
- Character AI's monthly active users fell from 28 million to 20 million between mid-2024 and early 2025, according to usage tracking.
- Character AI's valuation collapsed from $2.5 billion to roughly $1 billion, industry analysis shows.
- Character AI's filter interruption rate climbed to 23.5% by April 2026, comparative testing found.
- Roughly 75% of Character AI's subscription perks were locked to mobile apps, user reports revealed.
- Character AI's web traffic dropped another 13% between December 2025 and July 2026, per industry tracking.
- Replika's memory recall improved from 40% to 70% while Character AI's quality declined, comparison data showed.
- Under-18 chat restrictions in November 2025 triggered an 'emotional rupture' among users, industry analysis noted.
The Slow Unraveling: How Character AI Changed
Character AIβs evolution since 2024 has been marked by a series of incremental changes that collectively reshaped user expectations, leaving long-time users disillusioned. What began as minor adjustments to free-tier access eventually became a structural shift, with paywalls, restrictions, and moderation tightening over time. Industry analysis highlights that these changes βdid not arrive all at once,β but their cumulative effect eroded the platformβs original appeal.
Key shifts included paywalled swipes and βgo-ons,β which restricted free users from exploring multiple characters. User reports noted that free tiers faced chat limits, in-chat ads, and a βCharmsβ paywall, with 75% of subscription perks confined to mobile apps. By 2025, the platform imposed under-18 restrictions, removing open-ended chat for minors and triggering an βemotional ruptureβ among users who felt the βoriginal product was leaving them.β This move marked a turning point, as users described it as a betrayal of the platformβs early ethos.
Moderation also became more aggressive. By April 2026, the filter interruption rate rose to 23.5%, with users complaining that bots fell into repetitive patterns and βforgot detailsβ during conversations. Comparative data showed Replikaβs memory recall improved to 70%, while Character AIβs quality declined. These changes coincided with a 13% drop in web traffic between December 2025 and July 2026, according to industry tracking.
- Paywalled swipes and βgo-onsβ restricted free-user exploration
- Under-18 chat restrictions in November 2025 caused user backlash
- Filter interruption rate climbed to 23.5% by April 2026
- Subscription perks were largely mobile-only by 2025
A Reddit user summarized the sentiment: βCharacter AI feels like a house you built going up in flames.β As restrictions mounted, users migrated to alternatives like Candy AI and Janitor AI, which offered fewer limitations. Competitor growth reflected this shift, with platforms emphasizing unfiltered interactions.
FlirtAlly, with its real phone numbers, unlimited memory, and exclusivity features, emerged as a potential alternative for users seeking unfiltered AI companionship. Its approach contrasts with Character AIβs restrictions, positioning it as a haven for those prioritizing immersion over moderation.
The Numbers Behind the Exodus
The decline of Character AI has been a stark reality, with monthly active users plummeting from 28 million to 20 million between mid-2024 and early 2025. This downward trend continued, with a 13% drop in web traffic between December 2025 and July 2026. The valuation of Character AI also took a hit, sliding from $2.5 billion to $1 billion. Key factors contributing to this exodus include paywalls and feature restrictions, which limited free users' access to core features.
Some notable issues with Character AI include:
- 75% of subscription perks being locked to mobile apps
- increased moderation, leading to a sense of emotional rupture among users
- regulatory pressures, such as California SB 243, which forced the platform to adopt stricter safety measures
The mass exodus of users from Character AI has been driven by these changes, with many seeking alternative platforms that offer more freedom and flexibility. As users continue to migrate to other platforms, it remains to be seen how Character AI will respond to these challenges. In the meantime, platforms like FlirtAlly are positioning themselves as alternatives, offering unique features like real phone numbers and unlimited memory to attract dissatisfied users. With the AI companion market continuing to evolve, one thing is clear: users are seeking platforms that offer a more immersive and unfiltered experience. As the landscape continues to shift, it will be interesting to see how platforms like FlirtAlly capitalize on this trend.
Where the Users Went: The Scattered Alternatives
When a platform changes underneath its users, they don't leave in a single orderly line β they scatter. That's exactly what happened after Character AI's wave of restrictions, and the migration map tells a story of fragmented, unfinished searching rather than settled relocation.
The numbers behind the exodus are stark. Monthly active users fell from 28 million in mid-2024 to 20 million by early 2025, and web traffic dropped another 13% between December 2025 and July 2026. Meanwhile, the company's valuation collapsed from $2.5 billion to roughly $1 billion over the same stretch.
Each major alternative absorbed a different slice of the departing crowd:
- Replika β the memory-focused veteran, whose recall accuracy improved from 40% to 70% as it courted frustrated Character AI users (comparison data).
- Janitor AI β picked up users seeking looser content restrictions and a stronger free tier (platform guides).
- Candy AI β drew those prioritizing fewer filters and more adult-oriented companion experiences (the same guides note it as a top destination).
But here's the honest truth: none of these is a real home. As one industry analysis put it plainly, "no single tool is a perfect clone" of what old Character AI offered. Users scatter because every alternative solves one problem while ignoring the others β better memory here, fewer filters there, but never the whole package.