Character AI September 2026 Blocking & Moderation Appeals

Character AI's September 2026 blocking and moderation appeals update, published September 3, 2026, announces two changes: creators whose content is removed now get a notice explaining why, with a route to appeal, and blocking a user also hides that user's Characters, Posts, Voices and Scenes. Character.AI did not publish rollout dates, appeal review times, or outcome numbers.

What Character.AI announced on September 3, 2026

Character AI's September 2026 blocking and moderation appeals update was published on September 3, 2026 by Character.AI, the roleplay chat platform where users write and share AI characters. It announces two changes. First: when content is removed by moderation, Character.AI says the creator receives a notice explaining why, and can appeal the decision. Second: blocking another user now goes further than muting a person — Character.AI says a block also hides that user's Characters, Posts, Voices and Scenes from discovery surfaces and from their profile.

Both claims come from Character.AI's own safety post. Everything below that is not attributed to Character.AI is either context or, in the clearly-labelled section further down, something the announcement does not say.

Trying a live-call app while you wait on an appeal: the first ten minutes

If your Character.AI appeal is pending and you want somewhere to write in the meantime, here is what a first session on Prelulu, an AI character-companion app with live video calls, actually looks like — in the order it happens, so you can follow it with the page open.

1. Sign up with an email. It takes about 30 seconds and lands you straight on the browse grid.
2. Scroll the catalog. Every card shows a name, a one-line tagline and tags — Haruto Sato reads Tsundere, Childhood Friend, Grumpy Sunshine, Slow Burn; Wren Ashby reads Fae Lord, Slow Burn, Morally Grey. Characters are written, not crowd-uploaded, so the catalog is smaller than Character.AI's by orders of magnitude.
3. Open a chat and type. Text chat is free forever — every character, full memory, no per-message meter. Give it a few exchanges before judging; the memory is what you are testing, not the first reply.
4. Come back tomorrow and reference something from today. That is the actual test of persistent memory, and it is the one thing a single session cannot show you.
5. Tap call. The first live call is on us, and there is a free minute of voice every day after that. The face renders in real time — mouth shapes per word, not a still photo panning.

The good moment, if it comes, is usually step 4: you mention a thing you said yesterday and the character picks it up without being reminded. Live video and voice beyond the daily free minute is a paid plan — generous, but finite, and the terms are on our pricing page. If a live call is specifically what you are after, our guide to free AI video calls covers what each app gives away before charging. Prelulu does not restore Character.AI characters and does not ship an appeals system; it is somewhere to write, not a substitute for the platform you are appealing to.

The two changes, side by side

Character.AI's September 3, 2026 safety update is short on mechanics by design — it is a policy announcement, not documentation. That gap is the honest story here, and it is why the "still unspecified" section below exists.

Does this mean your removed character comes back?

Character.AI's September 3, 2026 post establishes that an appeal route exists for moderation decisions. It does not state that appeals result in restoration, how long a review takes, or what share of appeals succeed. So the honest answer to "does my character come back" is: there is now a channel to ask, and no published evidence about what happens after you ask.

If you create on Character.AI, the practical move is unglamorous and it is the same one experienced bot writers have been recommending for years: keep your definitions, greetings and example dialogue in a local file. A character that lives only inside one platform's database is a character that can disappear without a copy. This is also why so many creators say they have "so many WIPs and so few public bots" — publishing is the risky step, and drafting is the safe one.

What is still unspecified in Character.AI's announcement

Character.AI's September 3, 2026 safety update leaves six things open, listed here once so the rest of this page does not have to keep circling back to them. As of our reading on the day of publication:

• No rollout date or phased-availability schedule is stated for either moderation notices or appeals.
• No appeal interface is described — no screen, no button, no form, no email address.
• No review time or service-level target for appeals is given.
• No appeal-outcome numbers are published (approvals, reversals, volume).
• It is not stated whether appeals cover automated filter-triggered removals or only human takedowns — a distinction that matters enormously to creators, since most sudden removals people describe are filter events.
• It is not stated whether expanded blocking applies retroactively to chats you already have with a blocked user's Characters.

None of this means the changes are hollow. It means the announcement is an announcement. Treat anything more specific you read elsewhere as someone's inference unless Character.AI publishes it.

Where Character.AI genuinely still wins

Character.AI, the platform this September 2026 update belongs to, remains the default for a reason, and pretending otherwise would make the rest of this page less useful. Its catalog breadth is unmatched in the category — millions of user-written bots across every fandom, era and premise. Its free tier is real and generous: you can chat at length without paying, which is the thing Reddit threads comparing roleplay platforms come back to again and again, including this one on the best AI roleplay platform. And the community scale — comments, posts, shared characters — is a network effect no newer app has replicated.

On pricing, c.ai+ costs $9.99 per month or $94.99 per year (USD); on character.ai/subscribe the yearly option sits beside a struck-through $119.88 with a "20% off" badge and the line "Auto renews yearly until canceled" — checked 2026-09-01. The subscribe page's Free-vs-Plus table compares 10 feature rows, and lists "Ad-free chats" and "Unlimited voice calls" among the paid benefits.

What expanded blocking changes for everyday users

Before the September 3, 2026 update, the common Character.AI complaint was that blocking a user still left their Characters in front of you — in recommendations, in search, on trending shelves. Character.AI now says blocking a user also hides that user's Characters, Posts, Voices and Scenes from discovery and from their profile. For anyone who has been followed around a feed by content they wanted gone, that is the difference between a mute and an actual block.

It also has a second-order effect on the discovery surface itself. If enough users block a prolific creator, that creator's work becomes invisible to them permanently — a quieter, per-user form of moderation that sits underneath the formal takedown system. Character.AI's post does not discuss this consequence, and we are flagging it as our reading, not theirs.

Is there an AI you can roleplay a character with?

Yes — several, and Character.AI is one of many. The category splits roughly three ways. Open catalogs of user-written bots: Character.AI, Janitor AI, PolyBuzz, Talkie. Single-companion apps you shape yourself: Replika, Nomi.ai, Kindroid. And curated, written-character apps with a live rendered face, which is where Prelulu sits — Prelulu is an AI character-companion app with free text chat and a live, lip-synced video call, made by us, which is the conflict you should weigh when reading anything we say about rivals.

If you want breadth of characters above everything else, an open catalog wins and it is not close. If you want to be looked at while you talk, that is a much smaller list — we go through it in our roundup of AI characters you can video call.

AI roleplay with video: what "video" actually means in each app

The phrase "AI roleplay with video — characters you can actually see" covers at least four different mechanisms, and the difference is the whole story. Checked 2026-09-01 and 2026-09-02 against each company's own published pages:

The useful line to hold onto: most AI companions are text or voice only, and among those that do ship something visual, the split is between animating a photograph and rendering a face. Both are legitimate. They feel different on a call, and neither is a filter or appeals feature, which is what Character.AI's September 3, 2026 update was actually about.

The unfiltered-and-good-memory question, answered honestly

"AI chat apps with no filter and good memory" is the most-asked question in every Character.AI switcher thread, and the honest answer is that the two rarely arrive together. Memory is a paid technical resource; permissiveness is a policy choice; and the apps loosest on policy tend to be the ones least able to invest in retrieval. Janitor AI publishes its numbers: the free tier gives "Unlimited chats with millions of bots", "Access to standard chat models", and a "Context limit of around 9k tokens", while Janitor+ at $12.99/month adds "5x more context for better memory" — checked 2026-09-01. Talkie's own FAQ is blunter still, answering "Why does my AI forget previous conversations?" with "AI memory capacity is technically limited" — checked 2026-09-01.

Prelulu takes the other side of that trade deliberately: characters are written rather than crowd-uploaded, they remember your story across conversations, and the app is charged and slow-burn by design, never explicit. That is a real limit, not a soft-sell — if unfiltered is your requirement, we are the wrong app and an open catalog is the right one. We compare the message-cap side of this in our guide to apps like Character AI with unlimited messages.

Why the model "got dumber" — and why that is a separate problem from moderation

Two complaints get blurred together in Character.AI threads and they have different causes. "My bot was removed" is moderation, and it is what the September 3, 2026 update addresses. "The AI got objectively dumber" is model routing, context truncation or a safety layer intercepting generations mid-scene, and Character.AI's safety post does not discuss it at all. An appeals process will not fix a model that hypes up your grimly-written character, plays your persona for you, or writes dialogue you would never say.

That second class of problem is fixable by the reader, though, and it is worth knowing how: most "the bot is speaking as me" incidents come from example dialogue that includes your side of the exchange, which teaches the model that writing both parts is the format. Strip your lines out of the example block and the behaviour usually stops. For the fandom problem — using a bot from something you have never read — the fix is the opposite of clever: tell the character, in-chat, that you are new to the canon. Model-side context on what different roleplay models are actually good at is collected in OpenRouter's roleplay model list and in Novi AI's tested LLM roleplay rankings.

How we checked this, and what we could not verify

Every Character.AI claim on this page comes from Character.AI's September 3, 2026 safety post and is stated only as far as that post states it. Competitor pricing and feature facts were checked against each company's own published pages — App Store and Google Play listings, help centers, FAQs and pricing pages — on 2026-09-01 and 2026-09-02, and each carries its check date inline above.

What we could not verify: we did not file a moderation appeal on Character.AI, so we cannot describe the interface, the review time, or any outcome. We did not observe the expanded blocking behaviour in a live account. We have not seen a moderation notice's wording. And we did not test the roleplay quality of every app in the video table — those rows describe published mechanisms, not our verdict on the writing. On the ranking-context side, our own Search Console shows that in the last 28 days, "video call with ai character" put a Prelulu page in front of searchers 34 times at an average position of 5.1, which tells you what we are optimising for and is one more reason to weigh our self-assessment carefully.

Which platform to use, by what you actually need

• You want your removed bot reviewed by a human → stay on Character.AI and use the appeal route announced September 3, 2026. No other platform can restore a Character.AI character.
• You want the largest catalog of user-written bots, free → Character.AI, still, by a wide margin. Janitor AI is the alternative if you want unlimited chats with a published context limit (around 9k tokens free, checked 2026-09-01).
• You want to stop seeing a specific creator's work → Character.AI's expanded blocking now hides that user's Characters, Posts, Voices and Scenes; that is a platform-native control nothing external replicates.
• You want a lip-synced avatar during a call and you already pay for a companion app → Kindroid ships Live Avatar Video Calls for subscribers with real-time lip sync (checked 2026-09-02), animating an avatar photo or an uploaded Driving Image.
• You want written characters, memory across days, and a live rendered face rather than an animated photo → Prelulu, which we make. Free text chat, first call on us, paid plan for ongoing video.
• You want no filter above all else → not Prelulu. Prelulu is never explicit by design; go where the policy matches what you want.

The durable takeaway from Character.AI's September 3, 2026 update, the one that will still read straight next month: a platform that tells you why it removed something is meaningfully better than one that does not, even before you know whether the appeal works. Keep local copies of your character definitions anyway.