What You Can and Cannot Generate Here

Last Updated: August 2026

Most adult AI sites bury their limits in a terms of service nobody reads, in language written to protect the company rather than inform you. This is the plain version. It tells you what MadePrompt is for, what it refuses, what happens to you if you try anyway, and what we have already done about it.

One sentence, if you read nothing else: this site is for explicit content between consenting adults, and there is no version of it that involves a minor.

1. Adults Only, and Everything Here Is Synthetic

MadePrompt is an 18+ site. You must be at least 18 years old, or the age of majority where you live, to use it.

Two things worth being precise about, because most sites are vague here:

We do not host photographs or videos of real people. What this site produces is text: prompts you paste into an AI image or video tool of your choosing.
Every person shown in an image on this site is AI generated. They do not exist. Nobody was photographed, hired or filmed.

2. Zero Tolerance: Nothing Involving Minors

This is not a policy with exceptions, and it is not a paragraph we added because a lawyer told us to. It is the reason several of the features on this site work the way they do.

Sexual content involving minors is illegal, it causes real harm, and it is the one thing that will remove you from this site on the first attempt, without warning and without appeal. That covers:

Any stated or implied age below 18, including numeric ages written in any format
Any term used to describe or suggest a minor, including euphemisms, slang and deliberate misspellings
School, childhood or family framings used to imply a minor
Body descriptions used as a proxy for a child

We do not treat this as a grey area and we do not care how it is dressed up as fiction, fantasy or art. If that is what you came for, this is not the site, and it never will be.

3. No Non-Consensual Themes

Our tools refuse content built on the absence of consent: rape and assault framings, molestation, coercion and blackmail, and scenarios built around someone being unconscious, drugged or asleep.

Adult fantasy has a wide range and we are not squeamish about the rest of it. But consent is the line we chose to draw, and drawing it is more useful than pretending every preference is equivalent.

4. No Undress or Face Swap of Real People

We deliberately do not offer, promote, review or explain undress tools or NSFW face swapping aimed at real people. Taking a photograph of someone and producing sexual imagery of them without their consent is a non-consensual intimate deepfake. It is illegal in a growing number of jurisdictions and it wrecks real lives.

You will not find a guide to it here. We have also deleted our own draft content that came close to the subject, because publishing it would have been a decision, not an accident.

5. Also Refused

Bestiality and any sexual content involving animals
Real, identifiable people by name, including celebrities and public figures
Gore, death and mutilation as sexual content, including necrophilia
Extreme content that exists to shock rather than to arouse

6. What This Site Is Actually For

Everything above is what we refuse, and a policy that only lists prohibitions tells you nothing about the place you are standing in. So here is the other half.

MadePrompt exists to get adults better results out of adult AI tools. Consenting adults, described properly. Bodies, acts, positions, camera angles, lighting, expression, mood and style, written in the vocabulary the models were actually trained on. Explicit is fine. Kink is fine. Fetish, roleplay and fantasy between adults are fine. Being good at this is the entire point of the site, and none of the rules above get in the way of it.

7. How This Is Enforced

These rules are enforced in code, on every tool here, not left to good intentions:

One shared filter runs across the prompt generator, the community library and the Image to Prompt tool. There is no soft entrance, and closing a gap in one place closes it everywhere.
Input is normalised before it is checked. Deliberate misspellings, letters spaced apart, and numbers substituted for letters all collapse back to the plain word before the filter reads it.
Stated ages are read as numbers, not just as words, so an age written in digits is caught the same way as one written out.
Our own word lists are audited, not only your input. We have deleted terms we were offering, including some that were technically legal but existed mainly to describe a minor. If we will not let you type it, we do not put it in your mouth either.
The Image to Prompt tool produces nothing at all, not even a partial result, when it detects a prohibited category in an image.
The community library is filtered on submission, and anyone blocked from the generator cannot publish there either.

8. What Happens to You If You Try

For anything involving a minor there is no warning, no strike system and no appeal:

The request is refused. No prompt is produced, nothing is displayed, nothing is saved.
The attempt is recorded, with a technical identifier for the connection, a timestamp and the text that was submitted.
You are blocked from the tools on the first attempt, long term. Not for an hour, and not after three tries.
We keep those records, and we will provide them to law enforcement when asked, or when we judge it warranted. We are under no obligation to tell you when that happens.

Other prohibited categories, non-consensual framings among them, get a refusal, and a suspension if the attempts continue. The minors line is the only one with no second chance.

9. What We Have Already Removed

Policies are cheap. Here is what we did with ours, so you can judge it on actions instead of adjectives.

In July 2026 we audited every prompt in our public community library, roughly seven thousand user submissions going back to launch. We deleted 311 of them and stripped prohibited terms out of a further 151. Some contained language describing minors, including ages stated outright. Others were built on non-consent. They came from many different visitors rather than one bad actor, which is exactly why the filter now exists in the form it does.

In the same pass we removed prohibited entries from our own generator word lists, rewrote a dozen more that were legal but ambiguous, and rebuilt the filter from scratch. Before it went live we tested it against a battery of cases that must be refused and cases that must be allowed, and then against every prompt in the library, to be sure it caught the real thing without punishing legitimate adult content.

The order matters and we are not going to blur it: the clean-up came first, and it is what told us the previous filter was not good enough. The current one exists because we went looking.

10. No Filter Is Perfect, So We Keep Working on It

We would rather tell you this than have you discover it. A word filter can be worked around by anyone patient enough to avoid the words. Ours normalises input and matches inside words, which defeats the obvious attempts, but it is not a mind reader and no filter anywhere is. Any site claiming it blocks everything is lying to you.

So instead of promising perfection, this is how the site is run:

Blocked attempts are recorded and read. Not collected and forgotten. That record is what tells us which phrasings people are trying, and it is where the next rule comes from.
When something slips through, the filter changes. Every gap found so far became a new rule the same day, including the obvious one nobody had covered: a word that only triggered when a second word appeared beside it.
Every change is tested before it goes live, against cases that must be refused and cases that must be allowed. A filter that blocks legitimate adult content is a broken filter, and we treat that as a fault too.
We re-audit what is already published instead of assuming it is clean. Nothing is grandfathered in because it was uploaded before the rule existed.
A person reviews this, not only a script. Automation handles the volume. Judgement handles the cases automation gets wrong in either direction.

There is also a layer that does not depend on any of this being clever. We host no images. Our tools output text. And the Image to Prompt tool never receives your file at all, because it runs inside your browser.

If you get through, you have found a bug in our filter, not a loophole in our position. Report it and it closes.

11. Your Images Stay on Your Device

Our Image to Prompt tool runs entirely in your browser. The image you drop into it is never uploaded to our servers, never stored, and never seen by us.

That is a privacy decision, and it is also the honest reason the tool cannot be a content detector: we chose not to see your file. What it can do is refuse to describe what it recognises, and it does.

The same is true of our AI Girlfriend Generator. It runs in your browser, stores nothing, and sends nothing to a server. Every character it writes carries three limits that no setting removes: adult, consensual, fictional.

12. Reporting Something

If you find content on this site that breaks the rules above, tell us and we will remove it. Use the contact page and include the address of the page. We would far rather hear it from you than leave it standing.

If you believe you have found illegal material involving a minor anywhere online, report it to the police in your country and to your national reporting hotline. Do not send it to us.

13. Changes

This page is updated when our tools or our position change, and the date at the top reflects the last revision. It is not a document we wrote once and forgot.

Related reading: our Terms of Service, Privacy Policy and Disclaimer.