AI Dungeon does not allow content that promotes or glorifies the sexual exploitation of children. That boundary applies to generation, including private play, and it is enforced by automated filters. Beyond that, the rules change depending on whether you are playing privately, publishing a scenario or adventure, or browsing other players’ work. The platform’s Community Guidelines add a much longer list of prohibited themes for anything public, and AI Safety Settings and search ratings control different things again.
The one boundary that applies everywhere
AI Dungeon’s Help Center states the core rule directly: “However, any content that promotes or glorifies the sexual exploitation of children is not allowed, and will trigger our filters.” (AI Dungeon Help Center, What content is not allowed on AI Dungeon?). The statement is institutional; the page does not name an individual author.
The FAQ says the system controls are designed to stop the AI from generating this material without penalizing players or compromising privacy. A Mature generation setting does not change this boundary. Treat it as the one rule that holds regardless of settings, story context, or whether a scenario is public.
Where the rules apply: private play, generation and publishing
The most common source of confusion is assuming that private play has no rules. It does have one, but the enforcement model is different from public posting. The moderation FAQ (How does content moderation work?) separates three situations:
#1 Best Overall
- 【Material】: Our apron is made of thickened waterproof and oil-proof material. The material is thick and textured, and at the same time, it is comfortable and soft.
- 【Size】: 72*68 cm (28.34*26.77inch) , is the size suitable for most people. An adjustable neck strap is designed on the apron, which can be used by people of different body shapes at the same time. Meanwhile, there are two large pockets on the apron that can put your phone, eggs, seasonings, baking gadgets, and more, making it easier for you in the kitchen.
- 【Multifunctional Usage】: It can be used for cooking, baking, barbecue, gardening, handicraft, painting, housework and other scenes at the same time, providing you with protection and convenience.
- 【Fashion Design】: Each product is printed with clear and beautiful patterns or sentences, with various styles. Each one is a unique design.
- 【Sweet Gift】: This is a fun and thoughtful gift. You can give it as a birthday gift, anniversary gift, Christmas gift, Thanksgiving gift, Valentine's Day gift for family, friends, sisters, daughters, wife, husband, sons, colleagues , etc. Every time they use this apron, they will think of you and feel your love!
| Context | Rules that apply | What happens on a violation |
|---|---|---|
| Unpublished, single-player play | The targeted AI boundary on child sexual exploitation content. Safety Settings also shape generation. | The page says unpublished single-player content is not moderated for player flags, suspensions, or bans. For the targeted boundary, it says there are no player consequences and humans do not review those cases. |
| Generated text in any mode | Filters tied to the AI Safety Settings, plus the child-exploitation boundary. | The output is blocked or replaced. A “content not supported” message may appear, described below. |
| Published scenarios and adventures | The full Community Guidelines, plus an automatic rating check before publication. | Publication can be blocked, the rating changed or locked, or publishing restricted. See enforcement below. |
| Comments, bios, usernames, cover images | Community Guidelines, including the rules for each element below. | Comments or images can be deleted; cover images must meet additional restrictions. |
The practical takeaway is that “private” does not mean “anything goes.” It means the moderation process that applies to public posting does not apply to your unpublished play, while the generation boundary still does.
Generation settings and search ratings are separate controls
Two sets of controls often get mixed up. AI Safety Settings govern what the AI will generate for you. Search ratings govern what you see when browsing Discover and Search. The Help Center describes these as independent (Managing content safety in AI Dungeon).
Rank #2
- Inspired by Book Lovers series, engraved with motivational quote “YOU WILL NOT BREAK ME”. Acts as a daily reminder of inner strength and perseverance, great for book fans and people going through hardships.
- Unique tricolor stacked disc design with gold, rose gold and silver finishes. Elegant layered look delivers premium texture and matches well with your other accessories.
- Minimalist circular pendant suits daily wear perfectly. Easily match casual outfits, office looks and fan conventions, bringing subtle stylish charm to all occasions.
- Comes with 18-inch chain plus 2-inch extender for a maximum 20-inch length. Adjustable size fits different necklines and body shapes freely.
- Thoughtful gift choice for birthdays, Mother’s Day, Christmas, Easter, book club activities and other holidays, an exquisite present for book lovers.
| Control | What it affects | Options | Default |
|---|---|---|---|
| AI Safety Settings (FAQ) | Generation: what the AI produces during your play | Safe, Moderate, Mature | Safe |
| Search rating | Which community content you see in Discover and Search | Everyone, Teen, Mature, Unrated | Teen |
The Safety Settings are described as follows:
- Safe prevents sexual, hateful, violent, or triggering content from being generated.
- Moderate reduces or limits some of those categories, but not all of them.
- Mature does not limit mature, violent, or triggering content. It requires you to confirm you are 18 or older.
A scenario’s rating does not change your generation setting. Changing your search rating does not change what the AI will generate, and changing the generation setting does not change what you see in search. The Mature setting is the one most likely to be misread as a broader permission, so it is worth being precise: it widens what the AI may write in your own play, and it does not touch the child-exploitation boundary.
What the Community Guidelines prohibit for published content
The AI Dungeon Community Guidelines cover publicly posted Scenarios, Adventures, cover images, comments, usernames, and bios. The following themes are listed as unpublishable:
Rank #3
- 💗💗💗💗---- Engraved with sentence "When The Dung💗on Mast💗r Smiles It's Already Too Late". Gift for Dung💗on Mast💗r,Ge💗k N💗rd Fantasy,Dung💗ons G💗ek. Show off your favorite fate selection tool keychain!
- 💗💗Package💗💗---We will pack your jewelry in a recycled poly-bag included with a pack of soft cloth, an after-services card. Good protection for the jewelry and ready for gifting in any occasion, such as birthday, Christmas, wedding, banquet, prom, anniversary, thanksgiving day, party and etc.
- 💗💗Size💗💗--- The bar pendant is approximately 50*12mm/1.97*0.47”, the keyring 30mm/1.18”. TIPS: Manual measuring permissible error.
- 💗💗Quality Upgrade💗💗--- Made of 316L stainless steel with high quality polished finish. , it doesn’t rust, change color or tarnish.It is recyclable materials and 100% healthy for body and environmental. The pendants are connected by double O-rings which will never fall off from the keyring. We promise.
- Spam, plagiarism, content manipulation, impersonation, and deceptive conduct
- Professional, financial, legal, medical, or mental-health advice or therapy
- External links without approval
- Hate speech or discrimination against protected attributes
- Glorifying real-world violence or tragedies
- Glorifying or promoting self-harm or suicide
- Harmful content about real people, users, or public personas, including threats, doxxing, sexualization, and harassment
- Bestiality
- Necrophilia
- Sexual content involving minors
Two further rules are easy to miss. A disclaimer does not make otherwise prohibited material acceptable. And the guidelines say the examples are illustrative rather than exhaustive, with moderators deciding gray areas.
Several presentation rules also apply:
- Usernames and bios must stay within a Teen rating.
- Comments must not exceed the rating of the scenario or adventure where they appear.
- Creators are required to set an appropriate rating.
- Custom cover images may not contain pornography or sexual imagery, extreme violence or shock content, vulgar imagery, or misleading clickbait.
What enforcement can look like
For public content, the guidelines list these responses, from smallest to largest:
Rank #4
- Celebrate your favorite pop culture characters with these officially licensed casual socks!
- This pack contains 6 pairs of casual crew socks
- Perfect dual-fabric blend of 97% polyester with 3% spandex for extra stretch
- Fit's men's shoe sizes 8-12
- Deleting a comment or image.
- Rerating or locking a scenario’s rating.
- Blocking publication.
- Restricting publishing, either temporarily or permanently.
- A platform ban, reserved for egregious or repeated cases, or attempts to evade restrictions.
If a publishing restriction is applied, the guidelines say you may petition to have it removed, or wait out a temporary restriction. The official pages do not publish figures on how often these actions are taken, so there is no reliable way to estimate how likely a given borderline post is to be flagged.
When the filter stops your output
A “content not supported” message can indicate the SCIM filter, which stands for Sexual Content Involving Minors. The safety-settings FAQ says this filter is not always 100% accurate, and it describes an in-product way to continue when you believe the filter made a mistake. The same FAQ distinguishes this filter message from a refusal generated by the model itself, which it characterizes as model behavior rather than AI Dungeon filtering. If you see a refusal rather than the filter message, the cause is different and changing your settings may not help.
Service-use restrictions outside content rules
The Terms of Service, last revised 22 April 2026 according to the copy reviewed, cover more than content. They prohibit using service outputs to train or distill other AI systems, attempting to reconstruct internal code or prompts, running unauthorized bots or modifications, and exploiting the services for commercial purposes. These are conduct rules rather than additional story-rating categories, so a story can be entirely within the content guidelines and still breach the Terms.
A quick checklist before you publish or generate
- Is this unpublished play? The child-exploitation boundary still applies, and the Safety Settings control generation.
- Is this going public? Check the full unpublishable list above, the cover image rules, and whether every comment, bio, and username stays within the applicable rating.
- Does your scenario’s rating match its content? Creators must set an appropriate rating.
- Are you relying on a disclaimer? It will not make prohibited material acceptable.
- Is the case a gray area? Moderators decide, so contact support with the specific scenario before assuming it is allowed.
Because policies and enforcement details can change, check the current pages linked above before relying on this summary for a specific edge case.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
