Writing With an Uncensored AI: Start a Story, Keep It Going
Tense scenes and technical drafts are where assistants stop. Here is how to write with an uncensored AI: brief it first, work scene by scene, and verify.
Key points
- Hand the AI a short character sheet, the setting and a sample of your own prose before you ask for anything; the output drifts without them.
- Work one scene per request, and paste the previous scene's gist rather than asking for 'the continuation'.
- An uncensored model still gets facts wrong and still cannot reproduce an existing author's work: verify facts against sources and keep other people's text out of it.
Contents
You ask for the next scene and the reply is a refusal: the passage you wanted is too tense, too dark, or too specific. Or the draft stops halfway through a technical section. Both are the same problem, and both are what a no-refusal AI is for.
You can try it while you read: a no-filter AI runs in the browser, and the trial covers one text message without an account. What the model will not fix is the craft, which is where most of this article goes: how to brief it before you write, how to work scene by scene, and what to check afterwards.
What actually changes with an uncensored model?#
It stops refusing mid-scene. Arguments between characters, a killer's interior reasoning, a blunt medical conversation, or the mechanics behind a crime can be written to the end instead of cut off at the interesting part. Three things do not change, though.
- Accuracy. Historical and technical details still come back wrong sometimes. Check figures and proper nouns against sources.
- Copyright. Continuing someone else's novel, or imitating a living author's style, is a problem regardless of whether the model refuses.
- Responsibility. You are the author of what you publish or submit.
Other uses beyond drafting are in what people use uncensored AI for.
What should you prepare before you write?#
The model continues from what you give it, so with nothing prepared, the character's voice wanders between requests. Write a short brief first: who the characters are, where and when the story sits, and one paragraph of your own prose as a style sample.
Keep the character sheet short. For a site foreman named Megumi Miura, 42, the whole sheet might be: "Cautious, hears people out, keeps her feelings in. Speaks politely; drops into casual speech with people she is close to." For the setting: "a harbour town rebuilding after a typhoon, present day, deadlines set by the local authority." That is enough for the model to stop inventing a different person each time.
The style sample is your own writing, one paragraph pasted into the chat. The model picks up sentence length, punctuation and word choice from it. Do not paste another writer's work for this: imitating a specific author is not something you want in your manuscript.
How do you write scene by scene?#
Five steps, and you do not need them all at once.
- Ask for one line of direction. Start small: "Write a scene where a site foreman and a local fisherman argue about repair work after the typhoon, in an izakaya by the harbour."
- Add the brief. If that first scene drifts from your characters, paste the sheet and setting above and ask for the same scene again, this time working from the brief.
- One scene per request. Asking for the meeting, the argument and the reconciliation together invites stalling and skipped beats. Send them one at a time: "Now just the argument, in the site office, at night."
- Paste the previous scene's gist for continuity. "Write the next scene" drops the thread. Instead, give two or three lines of what just happened: "Megumi went quiet after hearing the delay, and Kazuya closed the drawings. Write the next morning on site."
- Let the sample polish the result. If the prose reads stiff, show the model your style sample again and ask it to rewrite that scene in the same voice, keeping the content and dialogue.
Treat everything that comes back as a draft. Read it, keep two sentences, cut the rest, and rebuild. For non-fiction, check every figure and name against the source before it goes anywhere.
What do you do if it still refuses?#
Change how you ask before you change services.
- Lead with the purpose: "I am writing this scene to show how the character's guilt works, so keep the focus on his reasoning."
- Ask for the general level: descriptions of atmosphere, decisions and consequences rather than step-by-step method.
- Shrink the scene: ask for one beat of it, not the whole sequence.
If it keeps refusing, the four reasons behind refusals explains what you are actually hitting, and waiting, adjusting or switching covers the alternatives.
What should you check before you publish?#
Four items, and the first two matter for anything with your name on it.
- Verify facts. Dates, names and numbers belong against primary sources.
- Keep other people's work out. Do not ask for continuations of published novels or pastiches of a specific author. Japan's Agency for Cultural Affairs publishes its thinking on AI and copyright for the general frame.
- Leave real people out of your inputs. Names, addresses and unpublished workplace details do not belong in the prompt; what to hand over and what to keep has the full list.
- Copy your work as you go. In the logged-out trial the conversation is not saved, so nothing is stored for you.
What is Deepent Research?#
Deepent Research is a browser-accessible LLM chat service built to refuse as little as possible. It keeps going with research that a standard AI would cut off mid-answer.
- It refuses nothing. We run uncensored and abliterated models with guardrails reduced as far as possible, so prompts a standard AI answers with "I can't help with that" still get a response (Qwen3.5 / Qwen3.8-27B-abliterated, depending on your plan).
- Your input is never used for training. What you look up is used only to provide the service, never to train AI or for any other purpose. Personal data is protected, and our privacy policy is public.
- Try it free. No GPU and no install, just open a browser. In the logged-out trial your conversation isn't saved, so you can start casually.
→ Try it free, no sign-up: https://deepentresearch.com
No question can stop research.
Frequently asked questions#
Can I use it without an account?#
Yes, for one text message, and the logged-out conversation is not saved. Signing up raises the number of messages, which is worth doing once you are drafting scenes in sequence rather than testing the water.
How long should the brief be?#
A few lines is plenty: two or three character traits, the place and period, the rule that constrains the plot, and one paragraph for style. Longer briefs make the model choose between competing instructions, and a short brief is easier to keep consistent across scenes.
Can I publish the text as it stands?#
Treat it as a draft. Check the facts, confirm it does not resemble an existing work, and edit it into your own voice before it goes out. The responsibility for a published piece stays with the person who published it.
What if it refuses a scene anyway?#
Lead with the purpose of the scene, move to the general level, or shrink the request to a single beat. If it still refuses, the refusal is about content rather than wording, and it is worth reading why refusals happen and what the alternatives are.
Disclaimer#
This article was drafted with AI assistance and then reviewed, fact-checked, and edited by the Deepent Research team before publication.
It is provided for educational and research purposes and is not intended to encourage unlawful activity or crime.
How this note was edited and verified
We cross-check primary sources such as official documentation and papers, and organize only what the facts support.
This note was drafted with AI, then fact-checked, expanded, and edited by the operator before publication.
Related notes
Why Does an AI Refuse? Four Reasons and What to Do Next
A refusal can come from four causes: a safety check, missing information, an unclear question, or a glitch or message limit. Read the message to find yours.
What Can You Hand to an AI, and What Should You Keep?
Summaries and background reading are easy to delegate. Personal data and publishing raw output are not. Here is where the line sits, and why, in five checks.
Can You Use an Uncensored Model Commercially? What to Check
Most uncensored builds trace back to Apache 2.0 or MIT, but the terms follow the base model and the distributor may add more. Check a model before you ship it.
Try it yourself
Deepent Research is an LLM chat service that refuses almost nothing. Everything covered in this note can be tried without downloading anything.