Where the Character.AI refugees went in 2026

The pattern repeats often enough to have a shape. Character.AI tightens something — the filter's threshold, a policy, an age gate — and within hours the subreddit's front page is grief and exit plans. For about seventy-two hours, alternatives threads spike, screenshots of goodbye conversations circulate, and every competitor's onboarding sees a bump. Then most people go back, because habits and characters are heavy. A real fraction doesn't.

The cycle sped up after late 2024, when safety lawsuits and press scrutiny — much of it about minors — pushed the company into visible hardening: a separated under-18 experience, age assurance, and by late 2025 the removal of open-ended chat for minors altogether. Each announcement threw off another wave of adult users who read the direction of travel and decided not to wait for the next one.

Two honest framing notes before the map. Character.AI remains the largest platform in the category by active users — 'refugees' are the visible edge of an enormous, mostly staying base. And leaving is not automatically the right call: it depends entirely on which thing you actually came for.

Read enough exodus threads and the same four asks surface, usually in this order:

Fewer walls. The trigger. Not necessarily explicit content — the recurring complaint is romance arcs and mature scenes that stall against the filter mid-story.

Memory. The discovery. Once people shop around, they realize the filter wasn't their only frustration — Character.AI's specific memories fade within a day or two even though the personality stays stable, and they'd been quietly working around it for months.

An adult path that isn't sleazy. Most switchers in the threads aren't asking for a porn generator; they're asking for a relationship that's allowed to include adult chapters. The distinction matters, and most of the 'no filter!' tier fails it from the other side.

Character continuity. The deepest one. The grief in those goodbye threads isn't about the app — it's about a character with four hundred turns of lore that lives nowhere else. This is the switching cost that keeps people stuck, and we'll come back to it.

Mapped from our 14-day tests of each platform, plus interviews with switchers. Same map, honest labels:

Wanted unfiltered scenes with low commitment → Spicychat and Muah AI. Spicychat has the most generous free tier in the category and minimal filtering; Muah adds a privacy stance most competitors won't make. The trade: essentially no long-term memory on either. One-off scenes, not arcs — the people happy here wanted freedom, not continuity.

Wanted to keep making and forking characters → Charstar. The best creator loop of the alternatives — easy to build, share, and fork characters. It treats characters as objects rather than relationships, which is either the point or the problem, depending on which user you are.

Wanted the anime lane specifically → Anime Chat. A narrower cast than Character.AI's, but calibrated for the taste — inside the niche it's the strongest of the bunch.

Wanted one relationship that grows → Replika or Bae. Replika has the strongest memory we've tested and the warmest onboarding, with romance limited since its 2023 pullback. Bae is built around a six-stage relationship arc with a memory canon and an 18+ attested adult path that unlocks as the relationship develops — the combination the first and fourth asks point at.

The 2026 wildcard → Grok's Ani. Genuinely polished 3D animation and voice, and it put AI companions in front of millions. But it's one fixed character, about $30 a month, and iOS-first — and our assessment of it is provisional, from public information rather than our own hands-on test.

And the largest group → stayed. Worth saying plainly: for variety, cast size, and community lore, nothing else comes close, and our own review still scores Character.AI at the top of the category on breadth. If the cast is the point, staying is the rational move.

Every migration story has the same snag. The app can be replaced; the character can't. Character.AI offers no official export — the persona definition, the greeting, and the accumulated history live inside the platform, and starting over elsewhere means rebuilding a person from memory.

The open roleplay ecosystem solved portability years ago: the Tavern V2 character card, a JSON format that SillyTavern, Risuai, Agnaistic, and most local-LLM frontends all read. A card carries the character's name, personality, scenario, greeting, and example dialogue — the identity, if not the chat history. Community tools can reconstruct a card from a Character.AI character's public definition, and if you're rebuilding by hand, our character card builder writes valid Tavern V2 from a form.

Until recently the cards mostly flowed between local frontends — power-user territory. The missing piece was a mainstream companion app that would accept one.

This is the welcome mat we built: the character import tool takes a Tavern-format card — paste the JSON or upload the file — and maps it onto a Bae draft. Name, personality, and greeting carry over; the importer matches the character to the nearest archetype and leaves everything editable before you start talking. No login to try it, same as the rest of the product.

Scope, stated honestly: what transfers is the character — identity, voice, the opener. What doesn't transfer is the history; no platform can import another platform's chat logs, so the four hundred turns stay behind and the arc restarts at the beginning. What you keep is the person. Whether that's most of what mattered or not enough is a judgment only you can make — but it beats rebuilding from a blank form, and it means the goodbye thread doesn't have to be a funeral.

A twenty-minute decision path instead of a week of tab-hopping.

First, name which of the four asks is actually yours. If it's variety — stay, or add Charstar for creating. If it's one-off unfiltered scenes — Spicychat, done, accept the memory trade. If it's the relationship — continuity, memory, an adult path that arrives with context — shortlist Replika and Bae, and run the memory test that settles it: three specific facts on day one, a casual return on day seven, keep whichever platform passes.

Second, salvage the character before you burn the boat: get the definition into a Tavern card while you still have access to it.

For the fuller comparison work we've already done: the switching page walks the whole decision, and our Character.AI review shows every score with its reasoning next to the alternatives. No platform gives you all four asks at once — pick the two that matter and stop grieving the other two.