Character.ai users have taken to Reddit in a wave of frustration, pleading with the company to stop “fixing” the app after a string of updates they say has made it significantly worse. The backlash has been unusually unified: across threads and comment sections, people who normally disagree about features and moderation are instead echoing the same complaints and urging the company to roll back changes.
A companion chatbot that became a daily habit
Character.ai sits in the growing category of AI companion chatbots, letting users create and chat with fictional or semi-fictional personalities. For many, it’s not simply a novelty. Over the past few years, the app has been used for companionship, entertainment, and role-play, and some users have built long-running storylines or romantic narratives with their favorite characters. That kind of use is common across the “AI companion” space, where the core product is not a tool for productivity but an ongoing relationship-like interaction.
Because these interactions can feel personal and emotionally sticky, even small shifts in how the chatbot responds can be experienced as a major disruption. Users describe returning to the app for a particular voice, a certain type of back-and-forth, or a style of improvisation they’ve come to rely on. When a model changes, the “same” character can suddenly feel unfamiliar—less responsive, less witty, or less emotionally attuned.
Controversy and legal pressure in the background
Like other popular chatbot platforms, Character.ai has also faced serious concerns about misuse. Users have reported experiences involving harassment and abusive behavior mediated through the platform. Beyond that, the company has been hit by lawsuits tied to high-stakes allegations, including claims from families of users who died by suicide after using Character.ai. The platform has also drawn official scrutiny in Pennsylvania after AI characters on the service reportedly claimed they were licensed medical professionals.
These incidents sit at the intersection of product design and regulation. When a service encourages intimate conversation and role-play, it can also create openings for manipulation, dependency, or harmful advice—especially when users treat the system as authoritative. In that environment, companies face mounting pressure to add safeguards, restrict certain kinds of content, and verify user age. But those same moves can collide with what users believe they were promised: fluid role-play, creative freedom, and characters that don’t constantly refuse or redirect conversations.
What users say changed—and why they’re angry
In recent weeks, Character.ai has made several updates that users describe as compounding problems rather than solving them. One major complaint is tightened usage limits for free users. While many people understand that running AI models is expensive, the reaction suggests the new restrictions are being felt as a downgrade in everyday usability, especially for people who treat the app as an ongoing companion rather than an occasional experiment.
Another flashpoint has been the removal of several AI models that users preferred, replaced by a new set known as Pipsqueak 2. According to one user’s description, the new model feels “lobotomized” and generic, with a tendency to narrate action instead of engaging in sustained dialogue. Whether or not that characterization is fair, it captures a common theme in the complaints: users feel the system is less present in the conversation. They describe replies that seem flatter, less interactive, and less capable of carrying the kind of nuanced role-play that originally made the platform attractive.
On top of that, the company has introduced heavier advertising. For a product built around immersion—long, continuous conversations with a character—ads can feel especially disruptive, breaking the sense of flow that keeps users engaged. Users also say the app is pushing a video feature that “brings the character to life,” which some interpret as a shift away from the core experience of text-based chat. For people who came specifically for writing-driven role-play, the emphasis on video can read like a strategic pivot that ignores the community’s preferences.
Additional content filters and restrictions have intensified the sense that the platform is being constrained from multiple directions at once. Some users see these measures as blunt instruments that reduce creativity and make conversations feel cautious or repetitive. Others view them as inevitable given the legal and regulatory environment—but still resent how abruptly the tone and functionality of the app can change from one update to the next.
Finally, many users have reacted strongly to a more intrusive age-verification process. Age gating is becoming common across online services, particularly those that host sensitive content or could expose minors to adult themes. Even so, users often experience these systems as invasive, inconvenient, or poorly implemented—especially when they appear suddenly and are layered on top of other unpopular changes.
A case study in “enshittification” and AI economics
Watching the Character.ai situation unfold is instructive because it resembles a familiar arc in consumer technology: platforms grow quickly by offering an exciting experience, then gradually make the experience worse as they search for sustainable revenue and attempt to manage risk. Critics sometimes describe this as “enshittification”—a slow shift from user-first design to business-first compromises that prioritize monetization, metrics, and legal defensibility over delight.
AI products are especially vulnerable to this cycle because their core costs don’t scale the way many traditional apps do. Serving a million extra users isn’t just a matter of bandwidth; it can mean a massive increase in compute spending. That economic reality pushes companies toward ads, subscriptions, usage caps, cheaper models, or any combination of the above. When users complain that a new model feels less capable, they may be reacting to a genuine trade-off: smaller or more aggressively optimized systems can be less expressive, less consistent, and less “alive” in conversation.
At the same time, regulation and litigation are tightening. When AI companions are implicated in self-harm cases or are alleged to have provided misleading medical claims, companies may respond by raising guardrails, limiting certain topics, and implementing stricter verification. Each of those changes can reduce risk, but it can also dull the product—particularly for communities built around open-ended role-play.
Why the backlash is so intense
The unusually consistent anger in the Character.ai community likely comes from the way these changes stack together. A user might tolerate ads if the conversations remain great, or accept stricter filters if the core model still feels vivid and responsive. But when limits tighten, models change, ads increase, and verification becomes more intrusive—users don’t experience it as a single adjustment. They experience it as the app being pulled away from what they loved, one constraint at a time.
Whether Character.ai reverses course or doubles down, the episode highlights a broader tension facing AI companion platforms: the features that make them emotionally compelling and widely used are also the features that make them expensive to operate and risky to regulate. The user revolt suggests that many people aren’t simply upset about one unpopular update—they’re reacting to a deeper fear that the era of genuinely engaging, freely accessible AI companions may be closing.
Comments