Does Janitor AI have a filter? What the rules allow

In its phone apps for iOS and Android, NSFW content starts switched off, which the help centre puts down to the store policies of Apple and Google.

The fix is official and simple on paper: once the setting is enabled through the web settings, the help centre says the app works exactly like the website.

I find it easier to think of the Janitor AI filter as a few separate layers:

• A content setting that you control, off by default in the mobile apps and switched on from the website
• Platform rules that no setting touches, including a zero-tolerance policy spelled out in the help centre
• An age rule: the service is meant only for people above 18, with an extra third-party check in three countries
• No price gate: adult content is a setting on Janitor AI, not a paid extra

My take: this is one of the fairer setups I cover. Plenty of companion apps treat mature content as the thing you pay for. Janitor AI makes it a free setting and puts the real limits in its rules, which I prefer, because you know where the line sits before you spend a cent.

What the filter is not: a word list you can argue with, or a switch you unlock with money. Inside the rules, with your setting on, the policy allows it. Outside the rules, nothing makes it acceptable, and I won't pretend otherwise here.

If you are moving across from Character AI and want to tidy up first, here is my walkthrough on clearing your old Character AI chat history. Further down, a comparison table puts Janitor AI next to two companion apps, so the content policies are easy to compare.

That clears up a common mix-up. A person installs the app, finds it tamer than the site, and assumes a new filter has arrived. Going by the official pages, it is the same platform with a different starting point on phones.

A few practical points from the official pages:

• The only official apps are the ones the help centre lists, and look-alike domains such as janitorai.ai or janitor-ai.org are not Janitor AI
• The content setting lives on the web, and the mobile app follows it once changed
• Buy Janitor+ inside the app and Apple or Google manages it; buy it on the web and Stripe does

My take: a tamer default on a phone is sensible. Phones get unlocked on trains and at work, and nobody wants a surprise there. What I don't like is that the explanation sits in a help article, so plenty of people never find it and assume the worst.

That point matters more than it first sounds. Two other apps on file, Candy AI and GirlfriendGPT, only allow adult content on paid plans. On Janitor AI, money buys memory and priority rather than permission, which I think is the honest way round.

One catch on the free side: the official wording promises unlimited chats, while outside sources describe a practical daily cap, so sources differ and neither number is one I'd build a routine on. My full Janitor AI review with the plans compared goes deeper on what Janitor+ is worth.

Here is how the content policies of the other apps on file line up, with the free start each one offers:

• Nomi AI: adult content allowed under its policy, and a free plan that never expires, allowing 50 messages daily at up to 400 characters apiece
• SpicyChat: adult content allowed, with free unlimited text chat on its Default model, plus ads and possible waiting lines
• OurDream AI: adult content allowed, with 55 Dreamcoins to start on the free plan at 1 Dreamcoin per message, although its help centre says it does not currently offer free dreamcoins
• Candy AI and GirlfriendGPT: adult content only on paid plans

My read of that list: if a free start with adult content allowed is the goal, SpicyChat and Nomi AI are the two I would open first. SpicyChat gives the most room for text chat on paper, and Nomi AI publishes an exact daily number, which beats a vague promise in my book.

Candy AI and GirlfriendGPT are a different bargain. You pay first, then the policy opens up. The trade works if images and voice together are what you're after, but I would not pick either one purely because Janitor AI's mobile default confused you. The web setting solves that for free.

One more split worth knowing: of the free-start options, only Nomi AI has web, iOS and Android apps on file, while SpicyChat and OurDream AI are listed as web only.

OurDream AI is worth a closer look if you want roleplay with group chat and media. Its free start is small and the sources disagree on it, so treat the Dreamcoins as a sample, not a plan. I have a longer run-down of apps that cover the same ground as Janitor AI if you want more options with their catches. OurDream AI's quick facts sit right below.

It helps to separate two things people mix up. Turning the content setting on through the web settings is the documented, allowed route for adults who meet the age rule. Trying to push a character past the platform's banned categories is something else entirely, and no setting, plan or outside model makes it acceptable.

What happens if an account goes? Janitor AI's help centre describes deleting an account as permanent: posts, characters, personas and profile information are removed for good and cannot be recovered. The record does not say what happens to data after a ban, so I won't guess. My point is simpler.

If you have spent weeks building characters and personas, that work is worth more than one scene that breaks the rules. A stalled scene costs you an evening. A lost account costs you every character you built, and that trade never pays off.

My take on the whole filter debate: Janitor AI already gives adults more room than most apps on file, for free. If that room still feels too narrow, the better move is a different app whose policy fits you, not a fight with this one.

The same question comes up elsewhere, and my answer does not change from app to app. I wrote about what the PolyBuzz chat filter does and what you can change for readers coming from there. For every other short question about these apps, from billing to settings, my help section with quick answers is the place to start.

One last practical note. If the web settings do not show what you expect, or the app still feels restricted after you change them, the help centre is the right place to ask. Sorting it out through official support keeps your account safe, which is the whole point.

For people connecting from Australia, Brazil or the UK, local law requires a one-time third-party age check through k-ID. The record lists three ways it can be done: scanning an ID, AgeKey, or facial age estimation. Nothing on file says the rest of the world gets the same check today.

My take: confirming your own age is a light check, and I would not be surprised to see more regions added over time, but that is a guess, not a record. For now, remember that the age rule and the content setting are two separate things. One decides who may use the service at all, and the other decides what you see once you are in.

If you live in one of those three countries, I would do the check before changing anything else. It is a one-time step, and it saves a confusing round of settings that seem to do nothing.

A different model changes the writing. It does not change the platform's rules. The record describes the adult-content permission as platform-wide, and nothing on file says a proxy or JanitorRouter moves the line on banned content.

So if someone sells a proxy as a way around the rules, I would treat that as a red light for your account, not a feature.

Where outside models do help is memory. The built-in JLLM keeps roughly the last 8,000 to 9,000 tokens of a chat before older parts slip away, and Janitor+ offers 5x more context. That is a quality question, not a filter one.

If you want to know where facts like these come from, my method page for these guides lays out the sources behind every page.

What is clear is deletion. Per the help centre, you go to Settings > Security > Delete account and approve it with a 6-digit code sent to your email, and the deletion is permanent. Posts, characters, personas and profile information are removed and cannot be recovered. I like that it is spelled out.

I don't like that the training question is still open. Save any character text you care about first, since nothing comes back.

My advice: whatever setting you pick, leave out your real name, your job and anything else that points back to you. That habit is free and covers the gaps. If discretion on your statement matters, the free plan has nothing to bill.

The catch is elsewhere. Janitor AI is text only, with no voice calls, images or video. Other reviewers report that free JLLM memory fades fast, with details slipping within 20-30 messages, which matters more for long stories than any filter does. If that is your worry, Janitor+ and its 5x context is the paid fix to weigh.

The price is a flat $12.99 a month and there is no token wallet, so at least the bill stays predictable.

If a friend asked me, I'd say: start free, flip the setting on the website, and spend a handful of evenings with it before you pay a cent. If you want pictures or voice, look at the other apps on this page instead of hunting for a way around the rules.

Coming over from Character AI? It runs a very different policy, and my separate page on Character AI's stance on NSFW covers that side of the story.