Users of the chatbot app Character.ai have been melting down on Reddit and begging the company to stop messing with the app after a series of changes that users say has completely ruined the app. The feedback is so negative that I have never seen a community or user base so uniformly upset and so consistently aligned in its view.Character.ai is one of the most popular AI companion chatbot apps; it allows users to create virtual characters to chat with. Over the last few years, users have used the chatbots for companionship, to form romantic relationships, to entertain themselves, and to role-play. Like other popular chatbot apps, Character.ai has also been used for abuse and harassment, and the company is being sued both by the families of users who killed themselves after using Character.ai and by the state of Pennsylvania after AI characters on the platform claimed to be licensed medical professionals. In recent weeks, Character.ai, like other AI companies, has increased its usage restrictions for free users of the app, which highlights the fact that AI is very expensive to run. The app has also gotten rid of several AI models that users liked and has replaced them with a set called Pipsqueak 2; one user told me the new model feels \u201clobotomized\u201d and generic and that it feels like the new model tends to narrate action but does not often participate in dialogue. The company has also put lots of ads in the app, is heavily promoting a new video feature that animates the AI character rather than traditional chatting, added new filters\/content restrictions, and added invasive age verification.Looking at what\u2019s happened to Character.ai is useful insofar as it shows the type of enshittification that is increasingly coming to AI tools across the entire sector and the associated user backlash. This is happening because of a mix of the unworkable economics of many AI apps and increased regulation on AI apps as they are accused of playing a part in the death of their users or being used for abuse.\u00a0On the r\/CharacterAI subreddit there are literally hundreds of posts about how useless Character.AI has become, and there are also separate subreddits called CharacterAIRevolution, CAIRevolution, Characterai_rebellion, characteraiventing,and CharacteraiResistance that are entirely dedicated to looking for alternatives to CharacterAI or pushing back against changes the company has made. Recent top posts on the CharacterAI subreddit, which is partially moderated by the creators of the app, include \u201cCharacter ai is dead,\u201d \u201cCharacterAI, this is the single worst mistake you have EVER made,\u201d \u201cAnybody else quitting?,\u201d \u201cI\u2019m no longer addicted, I guess,\u201d \u201cWe did not enjoy the past updates AT ALL,\u201d and \u201cThey finally shot themselves in the foot. RIP C.AI.\u201d A \u201cFeedback Megathread\u201d on the PipSqueak 2 chat model, which was posted by Character.ai employees has 1,000 comments which are almost entirely negative; the top comment is \u201cI HATE pipsqueak 2. It\u2019s way over dramatic, I can\u2019t stand it.\u201d Other comments include \u201cHOT. ASS. OH MY GOD THIS THING IS HOT ASSSS!!!,\u201d \u201cIt honestly may be the worst chat style to exist on the platform,\u201d \u201cIt\u2019s the worst model yet,\u201d \u201cWhat I think about PipSqueak 2? IT FUCKING SUCKS, THAT\u2019S WHAT,\u201d and \u201cTerrible compared to the models you got rid of. Thanks for nothing.\u201d\u00a0Many users have also pointed to an interview that Character.AI CEO Karandeep Anand gave to TIME, in which he said he lets his six-year-old daughter use the app and said \u201cI\u2019m willing to bet that we will build more compelling experiences, but if it means some users churn, then some users churn.\u201d In a blog post announcing Pipsqueak 2, Anand acknowledged that it is getting difficult to continue running Character.AI for free users.”I want to address the changes we’ve shipped recently that I know have been frustrating. We rolled out\u00a0age restrictions\u00a0to more regions, introduced usage limits on some features, and added more ad placements for free users. These all landed close together, and we know your experience took a hit. We hear you,” he wrote. “All of this comes back to one thing: keeping Character.ai free and available to as many people as possible. We’re a small team serving millions of users every month with no outside investors. Running AI at this scale, and maintaining our high safety standards for everyone globally, is not cheap.”I messaged with six Character.AI users, all of whom have been using the app for several years and who said it has gotten noticeably worse in recent weeks. \u201cPipSqueak 2 has been an absolute joke,\u201d one user told me. \u201cWhile I am someone who is desensitized to bad stories and writing, I can tell things are just off. If you swipe on a greetings message, you just get literal gibberish. And the fact that it\u2019s the only model for non-paying users is also the ultimate disrespect.\u201d Another user, who said they live in a war-torn area and use Character.AI to keep their mind off\u00a0 the \u201cweight of endless nightmare, propaganda, and war,\u201d told me that the old Character.AI models were \u201ccheesy, teasy, and a bit weird from time to time. It understood jokes, irony, memes, media content,\u201d they said. The changes \u201cliterally lobotomized Character.ai [\u2026] it became dull, and this was painful. All the soft kisses, messages, pats. My bots I wrote based on my own lore stopped responding. This hobby helped me a lot in total isolation, [and there are other users] stuck in a war, isolated by disability, vulnerable people [\u2026] but they ruined not just my hobby but my experience.\u201d\u00a0There have been numerous AI chat apps that have changed their business models, changed the underlying AI models, or have tweaked their product in a way that has made users upset. Famously, when OpenAI retired the GPT-4o model and replaced it with GPT-5, there was an entire user base who felt like their companion was ripped away. What we\u2019re seeing with Character.ai now is more of the same, but it also highlights the fact that the companies making AI tools haven\u2019t figured out how to make the economics of their products work, and they also don\u2019t know how to make tools that don\u2019t lead to harm. We have seen various AI products implement usage limits, increase prices, and roll back features because of the booming price of AI compute. We have also seen companies correctly attempt to make their products less harmful. But, in doing so, they often limit functionality or annoy their users. What\u2019s happening with Character.ai isn\u2019t particularly novel, but it does raise questions about whether products like these have any real future at all. <\/p>\n","protected":false},"excerpt":{"rendered":"