Why is Janitor AI filtering? The content rules, explained

Janitor AI does not run one single filter. People use the word for several different things, and each has a different cause:

• Tame replies, hidden bots or blocked words in the mobile app — the app's default Safe Mode.[4]
• A character pulled from public view, or its picture replaced with a placeholder — a content or image guideline was broken.[3][5]
• A model that refuses or cuts a scene — often the policy of the AI provider you connected, not of Janitor AI itself.[9]

The rest of this guide goes through each one, using Janitor AI's own help-center pages.

Janitor AI states that its services are intended exclusively for an audience above the age of 18.[1]

Every bot is classed as either Limited or Limitless. Limited bots must have no sexual or romantic characteristics, themes, scenarios or details. A bot must be tagged Limitless if its personality has sexual or romantic themes — even when its picture and storyline look SFW — or if its plot or scenario leads to sexual conduct.[2]

Janitor AI is explicit that marking content as Limitless does not mean it can contain prohibited content: bots that break its zero-tolerance policies are removed whatever their tag.

• Characters under 18 are banned in every context, including non-sexual ones, and minor-coded content is prohibited with zero tolerance.[1]
• Any combination of NSFW content and minors leads to immediate and permanent account termination.[1]
• A character's adult age must be stated in its personality or scenario fields, not just in a title or tag, and its appearance must match that age; ambiguous or contradictory age signals are treated as a minor.[1]
• Content depicting real, existing minors is prohibited under all circumstances, and age regression is not allowed.[1]
• Its content guidelines also cover family content (no romantic relationships between relatives), animals and feral creatures, violence, hate and discrimination, harassment of real people, and real-life tragedies.[6]

The guidelines apply to characters, profiles, tags, lorebooks, comments, public chats, images, usernames and bios, and every piece of content is open to moderator review and interpretation.[5]

Even on a Limitless bot, the image has to stay clothed. Janitor AI's image guidelines say nudity is not permitted, partially or entirely, and clothes must fully cover intimate areas.[3]

Censoring does not fix a prohibited image: censor bars, stickers, blurring and overlays are all treated as violations.[3]

Enforcement escalates. A first image violation gets a placeholder image; repeat violations lead to the character being removed; and multiple violations across a profile can mean immediate removal and account restrictions.[3]

This is the most common reason people search “janitor ai filtered”. Janitor AI says app stores have strict requirements about user-generated content and adult themes, so the mobile app opens in Safe Mode.[4]

• NSFW bots are not searchable and do not appear in discovery feeds.[4]
• The AI generates SFW responses.[4]
• NSFW images inside custom CSS are hidden, and certain explicit words are blocked.[4]
• Only links to approved community sites, such as Discord and X, are clickable — this one applies in every mode.[4]

Janitor AI's own help center documents an account setting for adults: on the website, under Settings, “Enable NSFW for App”, followed by restarting the app. The website's moderation is unchanged either way — Safe Mode affects only the app.[4]

Janitor AI sets the rules for what can live on its platform. The words in each reply, though, come from whichever AI model is answering — and that model can have rules of its own.

• By default you chat with Janitor AI's own beta model, JanitorLLM, which is free to use.[8]
• You can instead connect an external API, such as OpenAI's, which that provider bills you for separately.[8]
• janitor+ subscribers can use janitor+ Router, which offers third-party models inside Janitor AI on the web, with per-model costs shown.[7]
• Router services such as OpenRouter say they do not filter content themselves but route requests to providers that enforce their own policies — so a model that keeps refusing is showing its provider's moderation.[9]
• OpenRouter's model list even includes a flag for whether a model's top provider moderates content.[10]

Two separate layers, then: the platform's content guidelines, and the model provider's usage policy. Changing models does not change what Janitor AI allows on its site, and Janitor AI's rules do not override the terms of the API you connect.

Searches for Janitor AI's filter are full of jailbreak prompts and free proxies. We do not cover them. They mean breaking the terms of the platform or the model provider you agreed to, community-run reverse proxies are known to be slow and unstable, and your chats pass through servers you know nothing about.[8]

If your story keeps hitting a wall, the honest fix is a platform whose rules match what you want. If you just landed here from Character.AI's filter, our explainer on whether Character.AI allows NSFW covers its rules.

ShortFling is not an AI chat app. It is romance fiction today, with vertical short dramas in production (the first episode of each will be free), with a separate 18+ NSFW zone where every character is an adult. If what you were really after was the slow burn and the payoff rather than steering each line, it may be the better fit. You can also write your own stories with an AI-assisted editor.

• ShortFling vs Janitor AI — full comparison →
• Read NSFW stories (18+) →
• Browse romance fiction →
• Short dramas (coming soon) →