The creative work of a human being must not be replaced by the output of a machine.
Key Principles
Human Authenticity
Empathy, emotions, and human behavior can only truly be expressed by humans — even if machines are increasingly trained to imitate them.
Stories on OutaStory are written by people, for people. The richness and authenticity of human experience is what gives literature its power, and no algorithm can authentically replicate that.
Community Integrity
In our community of writers and members, real people should communicate with each other — not bots.
Every comment, review, and message on OutaStory comes from a real human being. We prohibit automated accounts and bot-generated interactions to protect the integrity of our community.
Responsible AI Use
A human must always make the final decision about a text, a solution, or a recommendation.
We recognise that AI tools can help structure complex tasks and improve efficiency. However, the final judgement — the meaning, the voice, the decision to publish — must always belong to a human being.
Practical Application
OutaStory uses AI selectively and transparently. So that you always know what is at work, this section names every place an AI system is involved and the specific services we use. Where a human makes the final call, we say so — and where nobody stands in between, we say that too.
Text
- Story descriptions — On request, Azure OpenAI (gpt-5) drafts a short description from the chapters of your story. It is a suggestion: you review it, change it, and decide what is published.
- Narration markup — So that a narration voice gets emphasis, pauses and pacing right, gpt-5 produces control markup and a short style profile of the story from the chapter text. The text itself is left untouched; a word-for-word check makes sure nothing is dropped or added along the way.
- Public-domain classics — For the imported public-domain works, gpt-5 writes a short description and a brief biography of the author and files the work into a category. These texts are published without human review; the stories themselves are unaltered and labelled as imported.
- Translations — English versions of our legal documents are pre-translated with gpt-5 and reviewed by a human before publication. Articles on our blog are machine-translated between German and English.
Images
- Cover artwork — When no artwork of their own is supplied, Azure OpenAI gpt-image-2 drafts cover candidates; a layout model (gpt-5) first plans where the title and name go. The author picks one — or uploads their own image. The default category covers and the image in our daily writing-prompt email come from the same model. A small number of catalogue stories nobody chose a cover for receive a generated one automatically each day.
- Marking — Every image we generate carries a machine-readable provenance marker in its metadata (IPTC "trainedAlgorithmicMedia"), as Article 50(2) of the AI Act requires.
Narration voices
- Text-to-speech narration — An optional audiobook version of published stories. The default is Azure Speech's HD voices; ElevenLabs is available as a labelled alternative when the author chooses it. The text remains the author's own work — only the voice is synthesised, and the player says so when a voice is AI-generated.
- Rewording for the speech filter — If the speech provider's own content filter rejects a passage, gpt-5 rewords exactly that passage so it can be narrated. This affects only the audio version; the published text is never changed by it.
Cloned voices
- Voice clones — Premium authors can have their own voice cloned at ElevenLabs (Instant or Professional Voice Clone) to narrate their stories with it. This happens only with the explicit consent of the person whose voice is cloned; for a Professional Voice Clone, a voice verification step at ElevenLabs additionally confirms that it really is their own voice. Background noise in the voice sample can first be removed with ElevenLabs' noise reduction. What happens to your data along the way is set out in the privacy policy.
Moderation and classification
- Safety screening — Azure AI Content Safety screens comments, chapters, cover images, cover text and writing prompts for hate, self-harm, sexual content and violence before they reach the community. Anything above a high severity threshold is rejected automatically; borderline cases go to human moderators, who make the final decision.
- Age rating of the classics — For imported public-domain works, gpt-5 proposes an age band (0/6/12/16/18) following the jugendschutzprogramm.de scheme; when in doubt the more cautious one is chosen, and a rating set by our team always takes precedence. Stories by OutaStory authors are age-rated by the authors themselves when they publish — no AI involved.
Assistant
- Ernest — Our voice and text assistant runs on ElevenLabs Conversational AI; its replies are generated by the Claude language model (claude-sonnet-4-6) from Anthropic, which ElevenLabs runs inside its own platform. It is labelled as AI, tells you so if you ask, and never writes into your story unasked: a suggestion only lands in the editor once you accept it.
Marketing copy
- Posts and notices — The captions of our social-media posts, the text of our daily writing-prompt email and the summary of new app versions are produced with gpt-5, in some cases without a human reading them before they go out. The writing-prompt email therefore carries an explicit AI notice.
What is not AI
Search and recommendations on OutaStory use no AI: search is a classic full-text search, and recommendations come from what other readers finished reading — transparent rules, not a learning model. Advertising and statistics do without AI as well.
In all cases, human creators remain at the centre of the OutaStory experience. AI is a tool in service of writers and readers — it does not replace them.
AI should help us become faster and better. It will not dominate us; it should serve us.
Write your stories without AI — for real people who read and listen to them.
Your Choices as a Reader
Stories whose author declared at publish time that they were created with AI are visibly labelled as AI-generated for everyone — on the card, on the detail page and above the text.
If you would rather not see AI-generated content, you can switch on the "Hide AI-generated content" preference in your profile settings. With it enabled, stories and covers produced with AI assistance are kept out of your feeds and recommendations wherever we can identify them.
Prefer a human voice?
Synthesised narration is always optional. For a fully human alternative, OutaStory's speaker marketplace lets writers commission professional human narrators rather than machine voices. Learn more on For Speakers, and see the Speaker Terms for the details.