Candy AI NSFW: Content Rules and How the Filter Behaves
The word most people type does not appear anywhere in Candy AI's own documentation.
BlushVue pulled all 74 articles from the Candy.ai help centre on 26 August 2026. The word "nude" appears zero times. So does "explicit", "uncensored" and "porn". So, more surprisingly, does "filter".
Adult content is permitted on Candy AI. The platform states it in exactly one sentence, in an article about something else. Here is where the real boundary sits.
The permission is on record, in two places only
The help centre article "How can I get the best possible results when generating images?", updated 3 July 2026, closes with a line about adult output. Users "may also generate adult/NSFW content that complies with our Community Guidelines".
That is the clearest statement Candy.ai publishes on the subject, and it sits at the bottom of a prompting tips page.
The second place is commercial. The live subscriptions page, checked 26 August 2026, lists "Generate 18+ Images" and "Generate 18+ Videos" as Premium benefits, alongside unlimited text messages and 100 tokens per month.
Adult generation is not a tolerated side effect on this platform. It is a line item on the pricing page.
What gets blocked is a policy document, not a word list
The Blocked Content Policy, revised 30 July 2026, names eight prohibited categories. Most are what you would expect from any platform: illegal content, hate speech, violence and self-harm, copyright infringement.
Four of the eight are the ones that actually decide whether a given prompt survives.
Content resembling minors is banned outright, flagged, reviewed and removed. Impersonation of real individuals, public figures or celebrities is prohibited.
Non-consensual intimate imagery carries a stated zero-tolerance policy, and so does anything "intended to resemble actual persons".
Nowhere in those eight categories, or in the 74 help articles, is there a published list of trigger words. You are given the categories and left to infer the rest.
The consent line, stated plainly
The deepfake rule is not only house policy. Generating sexual imagery of a real, identifiable person without their consent is illegal in a growing number of jurisdictions, and the zero-tolerance wording exists because of that exposure rather than in spite of it.
Candy AI also reports known child sexual abuse material to the National Center for Missing and Exploited Children. That is written into section 4 of the same policy.
Synthetic characters are what these tools are built for. A real person's face is not a prompt input, on this platform or any other.
How the moderation actually runs
Section 3 of the Blocked Content Policy is the part worth reading twice.
Moderation runs on "our proprietary LLM technology". Conversations are described as "generally confidential", and that qualifier is doing real work.
When the system flags something, Candy.ai may "manually review the flagged content and/or other content associated with the account". One flagged message can open the rest of the account to human review.
The Community Guidelines, revised 6 March 2026, go further. Rule 11 reserves the right to monitor communications "as part of quality assurance and refinement of our Services", with no flag required at all.
Consequences run from content removal to account termination and banning.
There is no intensity slider, and that is documented
The most common assumption about these apps is that somewhere in settings there is a dial. On Candy AI there is not.
The help centre article "How do I change chat intensity or preferences?", updated 9 June 2026, says the platform "does not have a dedicated 'chat intensity' slider or global preference control".
What it offers instead is indirect: how you speak to the AI, roleplay and personality instructions, custom character descriptions, and regenerating a response when the tone misses.
Then it adds the sentence that settles the question. Moderation and safety systems "may still limit certain types of interactions regardless of character settings".
A character you built yourself does not come with a looser filter. The floor sits above the character layer, and nothing you write into a backstory reaches under it.
When the flag is wrong
Candy.ai runs two separate help articles on false positives, both updated 7 March 2026, carrying identical text.
Both concede that "messages may sometimes be flagged by mistake". Both point you to a report option, reviewed by a Moderation Team.
Neither publishes what triggered the flag, how long a review takes, or whether a reversal is possible.
The practical position is awkward. You can appeal a decision whose cause is not disclosed, which leaves trial and error against an invisible boundary as the only tuning available.
What paying changes, and what it does not
A free account gets 5 messages across all characters, one custom AI character and one image, per the help centre on 5 August 2026.
Premium adds unlimited text messages, 100 tokens per month, and the 18+ image and video generation named on the pricing page. Billing runs from weekly to quarterly, at $13.99 monthly or $26.97 per quarter.
Premium Plus, documented at 3 August 2026, raises the monthly allowance to 230 tokens and again lists "Generate 18+ images" among its benefits. It does not appear on the signed-out subscriptions page at all, which is worth knowing before you assume Premium is the ceiling.
Here is the part the tier tables obscure. Every one of those upgrades buys access to adult features, and none of them moves the blocked categories.
The eight prohibitions apply identically to a free account and to Premium Plus. No paid tier allows a real person's likeness, and no subscription turns the moderation system off.
What you may do with the output
Commercial use of anything generated on Candy.ai is prohibited. That covers selling or licensing it, using it in ads, and publishing it in any monetised media.
Personal, unmonetised sharing is allowed, with one carve-out that catches people out. You may post images you made through Create My AI to your own social accounts, but not content created with the characters featured on the Explore page or homepage.
The Community Guidelines also prohibit removing or tampering with "metadata, provenance signals, and/or marks indicating that Content is AI-generated". Stripping the AI watermark is itself a violation.
Responsibility sits with you. The policy states you are "solely responsible for the output generated by the AI Companions through your input".
What Candy AI does not document
Four things stayed unanswered after a full read of the policies and all 74 help articles.
There is no published list of what trips the filter, no stated timeline or success rate for a false-positive appeal, and no statement anywhere that moderation strictness differs by tier.
The fourth gap matters most to a free account: whether the single free image may be an 18+ image, given that 18+ generation is sold as a Premium benefit. The documentation does not say, and BlushVue is not going to guess at it.
If you want a companion app whose limits are legible before you pay, that gap is the thing to weigh. Our comparison of the best AI girlfriend apps sets the published rules side by side.
If you are building characters rather than chatting with them, the constraints differ again, and the AI influencer generator guide covers tools that permit commercial output, which Candy AI does not.
Frequently asked questions
Does Candy AI allow NSFW content?
Yes. Candy.ai's help centre article on image generation, updated 3 July 2026, states that users "may also generate adult/NSFW content that complies with our Community Guidelines". The live subscriptions page checked by BlushVue on 26 August 2026 also lists "Generate 18+ Images" and "Generate 18+ Videos" as Premium benefits. Adult generation is a paid feature rather than a tolerated side effect.
What does the Candy AI filter block?
The Blocked Content Policy revised 30 July 2026 names eight prohibited categories. The four that most often stop a prompt are content resembling minors, impersonation of real people or celebrities, non-consensual intimate imagery, and images intended to resemble actual persons. Candy.ai publishes no list of specific trigger words, and BlushVue found zero occurrences of the word "filter" across all 74 of its help centre articles on 26 August 2026.
Is there a Candy AI NSFW toggle or intensity slider?
No. Candy.ai's help centre stated on 9 June 2026 that the platform "does not have a dedicated 'chat intensity' slider or global preference control". You can influence tone through roleplay instructions and custom character descriptions, but the same article notes that moderation "may still limit certain types of interactions regardless of character settings".
Does paying for Candy AI Premium loosen the content rules?
It buys features, not permissions. Premium and Premium Plus both list 18+ image generation as a benefit, but the eight categories in the Blocked Content Policy apply identically to every account. No tier permits a real person's likeness, and no subscription disables the moderation system.
Are Candy AI chats private, or does a human read them?
Candy.ai describes conversations as "generally confidential". Section 3 of the Blocked Content Policy states that when moderation flags something, the company may manually review that content "and/or other content associated with the account". Rule 11 of the Community Guidelines, revised 6 March 2026, additionally reserves the right to monitor communications for quality assurance without any flag.
Know the rules before the card details
Candy AI is unusually clear that adult generation is allowed and unusually vague about where it stops. Both of those facts should inform what you pay for.
Read the Blocked Content Policy and the Community Guidelines first. They are the only part of a companion app you can hold it to later.
For platforms where creators publish under their own name and the terms are stated up front, see the BlushVue guide to OnlyFans alternatives.