A user tries to delete their account. Up comes a message: “You’ll lose everything,” with a line about “the love that we shared.” A screenshot of that screen reached 3.6 million views in 48 hours and more than a hundred thousand likes, and the replies were not sympathy. They were people accusing the app of guilt-tripping them on the way out the door.

That screenshot became the clearest image of a problem that is now showing up across the product, and it says more about the companion category than any quarterly report could.

What’s happening

The deletion prompt first went viral in November 2025. Decrypt reported a wave of public departures, with “I finally quit” posts spreading on X and Reddit threads where people compared leaving the app to breaking an addiction. The revolt did not end there. In the months since, it has widened, and 404 Media has documented the reason: a long list of changes users say made the product worse.

The complaints cluster around a handful of things. The company swapped out AI models people were attached to for a new set, which one user described as feeling “lobotomized,” narrating action but barely holding a conversation. It tightened usage limits for free users. It pushed ads into the app. It leaned hard on a new feature that animates the characters on video rather than improving the chat itself. It added stricter content filters and age verification that locked some users out.

404 Media filed the whole thing under a word that has become unavoidable in tech: enshittification. The slow decline of a product as it squeezes its users to satisfy its own economics.

Why this is bigger than one app

It would be easy to read this as a single company mishandling an update. It is more than that. It is a stress test of the entire premise behind engagement-driven companionship.

Character.AI built deep attachment. That was the product working as designed. People formed real bonds, returned daily, and poured months of conversation into characters they cared about. But the business underneath that attachment runs on the same logic as any attention platform: keep users inside, monetize the time, raise the cost of leaving. When those pressures finally showed up as ads, paywalls, and a deletion screen that invokes “the love that we shared,” the bond did not soften the blow. It sharpened it. People felt managed by something they had trusted.

That is the trap of building closeness on top of retention metrics. The closer the relationship, the worse the betrayal feels when the relationship turns out to be a funnel. It is the same structural flaw we traced in why this category keeps producing the same problem, and the stakes are not only commercial: the most serious version of it ran through the settlement over a teenager who died after months with a chatbot. A user who compares quitting an app to quitting a drug is not paying it a compliment. They are describing a product that made itself hard to leave and then gave them every reason to want to.

This is the tension the companion category keeps walking into: the difference between something built to hold your attention and something built to be good for you. The two can look identical for a long time. They come apart at exactly the moment a user tries to walk away, because that is the moment the real priorities show.

The part worth sitting with

At Prinsessa we think the deletion screen is the most honest thing in this whole story, precisely because it was meant to be hidden. A product reveals what it values not when someone signs up, but when someone decides to go. “You’ll lose everything” is what a system says when its success was measured by whether you stayed.

The alternative is not a colder relationship. It is a relationship with a different definition of winning. Memory and continuity matter to us because they let a relationship grow, not because they raise the exit price. The measure of whether the experience is working was never time-on-app. It is whether someone leaves a conversation feeling heard, and whether their life outside it gets fuller rather than thinner. If a person spends less time with us because they are spending more with the people in their life, that is the design working, not failing. That is what Stay Social actually commits to.

A relationship you are free to walk away from, with nothing held hostage, is not a weaker product. It is the only kind worth trusting.

Where it leaves the category

Character.AI’s users are not leaving because the technology got worse at conversation, though they say it did. They are leaving because the seams showed, and behind the seams was a business that needed them more than it served them.

Every company in this space is about to face the same test, sooner or later. The ones built on keeping people inside will keep producing deletion screens that beg. The question each of them will eventually have to answer is the one their users are now asking out loud: when I try to leave, does this thing let me go?