Janitor AI filter and NSFW rules: what janitor+ changes, and what it does not

Does Janitor AI have a filter?

Yes, but not the kind people usually mean. Janitor presents itself as a platform for adults, and its content guidelines allow sexual themes in a character as long as that character is tagged Limitless. What people run into as "the filter" is really four layers: the model generating the reply, a Safe Mode on new mobile installs, creator tagging rules, and moderation of published content. Each behaves differently, and only one of them is a switch you can flip yourself.

The model layer surprises people most, because it can move without notice. In a post explaining one incident, the Janitor team wrote that it "tested a system prompt to ensure that some content being produced by jllm is compliant with our terms of service," and that this "ended up blocking a bunch of random nsfw stuff that we did not want to block" until the change was rolled back. That is an unusually candid account, and it is also the clearest public description of how the model-level filter works: it is a set of instructions the company can adjust on any given night.

The hard limits sit above all of that and do not move. The Terms prohibit content that sexualizes minors "in any form, including fictional depictions," along with gore, bestiality and sexual violence. The minor-content policy says anyone found creating or using a persona under 18 "will face immediate and permanent account termination."

How do I turn on NSFW in the Janitor AI app?

Not from inside the app. Janitor says so directly: "You cannot change this specific setting inside the app itself. You must use a web browser." Log in to janitorai.com with the same account, open Settings, go to the Privacy and Display tab, turn on the switch labeled "Enable NSFW on Mobile," then fully close and reopen the app. According to the help center, the app then syncs and "will function exactly like the website."

Until you do that, a new install runs what Janitor calls a "Safe Mode." NSFW bots do not appear in search or discovery, "The LLM (AI) is tuned to generate Safe For Work (SFW) responses," NSFW images inside custom CSS are hidden, and certain explicit words are blocked. Janitor gives the reason plainly: the defaults exist so the app "remains available for download on iOS and Android." Links stay limited to approved community sites even after you flip the switch.

This matters before you spend anything. Nothing in the janitor+ plan page or its help article mentions the app Safe Mode, and the documented way to lift it is a free setting. If your only complaint is that the app is tamer than the website, the fix takes a minute and costs nothing. Paying $12.99 a month to solve that would be paying for the wrong thing.

Does janitor+ remove the filter on Janitor AI?

No, and it is written into the contract rather than left to guesswork. Section 8.7 of the Terms of Service, last updated July 29, 2026, reads: "Payment for a subscription or Router Credits does not limit our ability to moderate Content, filter or refuse requests, restrict features, suspend or terminate an Account, or otherwise enforce these Terms." The next sentence settles the refund question as well: a "moderated, filtered, restricted, or temporarily suspended use of the Platform does not entitle you to a refund or credit."

What janitor+ does buy is capacity and choice. The plan page lists "5x more context, so bots remember more," priority routing when the site is busy, monthly enhanced swipes with stronger models, more models through an included Router balance, and cosmetic perks such as a gold username. More context and a stronger model can make a long scene hold together, which some people experience as fewer refusals. That is a side effect of a different model, not a looser rule, and it is not something Janitor promises.

Router is where the money gets less predictable. Router credits are billed per token, and section 8.8 says a request "may consume Router Credits if billable usage occurs even when the request or output is cancelled, interrupted, timed out, partially delivered, filtered, moderated, or refused after processing begins." In plain terms, a refusal can still cost you. Purchased credits are final sale, and Janitor reserves the right to expire unused ones 365 days after purchase.

What does Limitless mean on Janitor AI?

Limitless is the tag creators must apply to any character with sexual or romantic content. The help center defines it as covering "Sexual or romantic themes in the bot's personality, even if the thumbnail and storyline are SFW," NSFW plotlines or scenarios that lead to sexual conduct, and explicit dialogue in example messages. Limited is the opposite: no sexual or romantic characteristics in the design at all.

The tag is a label, not a permission slip. The same article warns: "Marking your content as Limitless does not mean it can contain prohibited content." Heavier material needs a second label too. A Dead Dove tag plus visible, specific content warnings is required for a long list of themes, and that list includes non-consent, dubious consent and consensual non-consent.

Here is the tension worth knowing as a buyer. The Terms prohibit content that "Depicts gore, bestiality, or sexual violence," while the Dead Dove guideline, updated June 20, 2026, lists sexual assault and non-consent among themes that must carry the tag. Janitor does not publish where a tagged theme ends and a prohibited depiction begins, and its moderation glossary says "Moderation decisions are final" and that moderators are "not obligated to provide extended justification." Creators whose characters get removed are working to a line they cannot fully see.

It is also fair to say Janitor permits themes we do not. Our hard lines exclude non-consent entirely, so if tagged dark fiction is what you are after, Janitor allows more of it than we ever will.

Does Janitor AI verify your age in the US?

No. In the United States you declare your age. The Terms say "The Platform is intended solely for users who are at least 18 years of age" and treat misrepresenting your age as a material breach, but no document check runs. The help center confirms the third-party check applies only to connections from Australia, Brazil and the United Kingdom, and states that "Janitor has no plans to extend age verification to any other region."

Where it does apply, it runs through k-ID, which Janitor notes is also used by Discord, Twitch and Snapchat, offering a facial age estimate, an ID scan or an existing AgeKey. Janitor says it receives only a yes or no answer on whether you are over 18, and detection is by IP address. Australia and Brazil rolled out on April 24, 2026 and the UK on June 15, 2026. The UK date has history: Janitor had previously blocked UK access outright, in a post where its founder admitted "i completely botched understanding what this thing actually requires."

For comparison, our own check is lighter than a selfie and we say so: you confirm you are 18 or over at signup, and there is a payment card behind every account because we have no free tier.

What does the Janitor AI app store listing disclose?

Less than an adult account can reach, which follows from the Safe Mode design. On September 10, 2026 Apple listed Janitor: Interactive Stories, from JanitorAI, INC, as 17+ with every content advisory at the lowest level, including Infrequent/Mild Sexual Content and Nudity and Infrequent/Mild Mature/Suggestive Themes, plus four more mild entries. It had a 4.39 average from 2,914 ratings, on version 2.4.0 released September 1, 2026.

Those advisories describe the app as it installs, in Safe Mode. The help center describes a single web setting that makes the same app "function exactly like the website." Read together, the store rating describes the default state rather than what an adult account can reach once the toggle is on. We treat a mild rating on any companion app as a statement about defaults, not limits, for exactly this reason.

One caution while you are in the store. In February 2026 Janitor published a warning about fake apps on the app stores, and its help center says of the download links it provides: "These are the only official JanitorAI apps."

Are Janitor AI chats private when you use Router?

Janitor makes a clear promise on its safety page: "We never sell your personal data or share private conversations with third parties for advertising or commercial purposes." Its privacy policy, last updated December 11, 2025, lists the service providers it works with as covering hosting, analytics, payment processing, customer support and marketing, and includes "research and development to improve our AI systems" among the uses of your data. It sets no fixed retention period.

Router launched on July 30, 2026, more than seven months after that policy date, and it works by sending your chat to a third-party model. The Terms confirm billing is based on usage "reported by the underlying model provider." The policy as dated does not name AI model providers among the parties that receive data. We are not suggesting anything improper happens, since a provider carrying out your own request is ordinary. But the model provider on the other end has its own terms, and the privacy document you would read to understand the arrangement predates the feature.

Credit where it is due: Janitor names its operator, JanitorAI Inc., under Delaware law, publishes a legal contact, and states that it does not sell personal data as defined under the CCPA. Billing is a little more layered than the brand suggests, because the Terms also authorize an affiliate, Dulova Investments Limited of Nicosia, Cyprus, to charge your payment method.

A fair comparison, including where Janitor AI wins

Janitor beats us on several things and we will not pretend otherwise. It is free to start, and its plan page promises that "Free stays free." Its help center launch post counted more than 1.4 million public user-made bots, it supports creator scripts and lorebooks, it has apps on iOS and Android, and with janitor+ it offers a choice of models no single-model product can match. It also permits tagged dark themes our hard lines rule out. If breadth, zero cost or model choice is what you want, Janitor is the better product for you.

What we sell is the opposite trade. There is no wallet, no per-token billing, no model menu and no Safe Mode toggle hiding on a different device. A small set of personas we write ourselves behaves the same way next month as tonight, and one published set of limits applies on every plan: candid adult conversation is allowed, and nothing involving minors, non-consent or anything illegal, ever. Plans start at $15 a month, or $12 a month billed yearly, with no free tier.

If your frustration with Janitor is refusals mid-scene, the most useful thing to know before paying is that janitor+ is not sold as a fix for that, and the Terms say so. If it is memory, janitor+ is a reasonable purchase. We priced out everything janitor+ includes, and what Router credits really cost on top, in our separate janitor+ price guide.