Publication & Review

Last updated: 2026-10-04What happens when you publish a character or world — publication is immediate, and review happens afterwards, only when something is reported.

Publication & Review

This document describes what actually happens when you publish a character or world to ASAI's public library, when content gets reviewed, and what we can do about it.

The short version: publication is immediate and nothing checks what a work says first. Review is reactive — it happens when somebody reports something.

Effective: 31 July 2026 · Version: 2.3
Operator: APERTURESyndicate OÜ (Republic of Estonia (EU), Registry code 17384111)

1. Two states: private and published

Every character or world you create starts as private. Private content is visible only to you. It is not reviewed, not scanned, and not seen by us unless you report a problem with it or a lawful order requires us to look.

Clicking "publish to library" makes it published. Published content is visible in the public library, discoverable by search, and usable by anyone.

You can unpublish at any time. The character or world stays in your account as a private draft. Other people's chats started from it end: their next turn is refused with a note that the work was unpublished, and the text already played stays with them.

2. What happens when you publish

Four things:

  1. We check that the definition is within the size limit (Worlds & Characters, Section 3)
  2. If the work is going into the Discover list, we check that your recent publishing does not look like an automated flood (below)
  3. We publish it — any account can, including Free
  4. A world's scenario openings are translated into the languages ASAI supports

That is the entire process. It usually takes seconds.

The flood check. It looks at the pattern of your publishing over the last day, not at what the works say: many works within an hour, works that are near-copies of one another, publishing at an even machine rhythm, works published seconds after they were created, near-empty cards without images. Real work — long, varied text with images — counts in your favour. If the pattern looks automated, publishing to the Discover list is held for a limited time (the app shows until when), and the account is flagged for a person in support, who can lift the hold early. Nothing already published is removed by this check.

2.1 What does not happen

An earlier version of this document described a pre-publication review system in detail: a classifier checking each field, a visual classifier on uploaded images, a "Pass / Soft flag / Hard block" outcome, a human review queue with service levels, and automated monitoring of published characters against a crisis-flag rate.

None of that exists. It was written as a description of an intended system and published as though it were a description of a working one. To be explicit:

  • No automated check judges the name, description, personality, history, greeting, tags, or universe field against our rules
  • No classifier examines uploaded images
  • Nothing is queued for human review at publication time, except an account flagged by the flood check
  • Published characters are not monitored, sampled, or spot-checked
  • There is no crisis-flag rate: the crisis check in chats (Crisis & Safety Resources, Section 8) keeps nothing about the character and is not used to judge it
  • Characters that break the rules — including prohibited archetypes such as therapist personas, and universes on the no-publish list — will publish successfully

The rules in the Content Policy still bind you. They are simply not enforced at the moment you press publish. They are enforced afterwards, when we find out.

3. How review actually happens

Review of what a work says is triggered by one thing: somebody tells us.

That means a user reporting content, a rightsholder sending a notice, an authority sending an order, or one of us noticing something while using the product. There is no queue of unreported works, no sampling, and no automated flag on content.

You can report in the product or by writing to us:

We accept reports in English or Estonian and aim to acknowledge within 72 hours on working days. Anything involving a minor, a credible threat, or imminent danger is prioritised over everything else.

Please include a link to the character or world and say what is wrong with it. A report that just says "this is bad" is hard to act on.

4. What we can do about it

When we review reported content, one of these follows:

4.1 No action

The content is within the rules. The report is closed. We do not notify the creator that a report was made about them.

4.2 Change requested

The content needs edits but is not a serious breach. We contact the creator explaining what has to change. If it is not changed, we unpublish it.

4.3 Unpublished

The content is removed from the public library and reverts to a private draft in the creator's account. The creator is told which rule was breached and what we did — this is our obligation under Article 17 of the EU Digital Services Act.

4.4 Deleted

For content that must not exist on our systems at all — child sexual abuse material, content that is otherwise unlawful — we delete rather than unpublish, and preserve or report as the law requires.

4.5 Account action

For serious breaches or established patterns: a warning, loss of publishing rights, suspension, or permanent termination. Enforcement follows the ladder in the Acceptable Use Policy, Section 9.

4.6 Hidden from the Discover list

The work stays published and opens by link, from players' shelves and inside worlds, but no longer appears in the Discover list or in its tag suggestions. This is a restriction of visibility, and you are told about it in the same way as an unpublication.

4.7 Published copy corrected

Where only part of a work breaks the rules, we may correct the published copy — for example, remove one offending passage — instead of taking the whole work down. The correction is recorded as ours, your own copy in your library is not touched, and your next update of the publication replaces it. You are told what was changed and why.

5. Bad-faith reporting

We treat reports as good-faith signals. Where someone repeatedly files reports that are plainly unfounded, or uses reporting as a way to harass a particular creator, we may stop accepting their reports and, in serious cases, take action on their account. This is the measure against misuse contemplated by Article 23 of the EU Digital Services Act.

Honest mistakes are fine. Nobody is penalised for a report that turns out to be wrong.

6. What we tell you, and what we don't

When we act on your content, we tell you the specific rule, what we did, and how to reach us about it.

We do not tell you who reported you, and we do not share information about other users involved. This protects reporters from retaliation.

7. If you think we got it wrong

Reply to the message we sent you, or write to [email protected]. A human reads it and can reverse the decision — content gets restored and account actions get lifted when we were wrong.

We are not going to promise you a formal appeals process. APERTURESyndicate is a micro enterprise, exempt under Article 19 of the EU Digital Services Act from the internal complaint-handling obligation in Article 20. An earlier version of this document promised acknowledgement in 72 hours, review by a different person, escalation to a second reviewer for close cases, and a written decision within 14 days. We cannot deliver that, and we should not have written it.

Your right to take a dispute to a court or a competent authority is unaffected. See DSA Notice & Action.

8. Special cases

Rightsholder objection without a formal notice

A rightsholder who contacts us informally about a published character may get it removed without filing a formal notice, at our discretion. We tell the creator when this happens.

Repeated problems across several characters

If several of your characters raise the same issue, we act at the account level rather than removing them one at a time.

9. Changes to this policy

Material changes are announced by email, by in-product notice, and through the date and version history on this page.

Version history

  • 2.3 — 4 October 2026 — Section 1: unpublishing ends other people's chats started from the work instead of leaving them unaffected.
  • 2.2 — 1 October 2026 — Section 2 corrected: any account can publish, and two automatic checks run at publishing — the size limit and a check for automated publishing floods, which can hold publishing for a limited time and flag the account for support; neither judges content. Section 3 lists the Report item and the report form. Two outcomes added: hidden from the Discover list (4.6) and published copy corrected (4.7).
  • 2.1 — 5 September 2026 — Team description corrected: the company has no permanent staff beyond its two founders, and around twenty volunteers contribute irregularly. Response expectations are unchanged — they are still written for the two people who are always on, not for the wider group.
  • 2.0 — 31 July 2026 — Rewritten to describe the actual process. Removed the pre-publication classifier, the field-by-field checks, the visual classifier, the Pass / Soft flag / Hard block outcomes, the human review queue and its service levels, the random sampling, the continuous post-publication monitoring, and the 5% crisis-rate trigger — none of which exist. Publication is immediate; review is reactive. Replaced the appeals process with an accurate description of what we offer.
  • 1.0 — 16 May 2026 — Initial publication

Contact

Operator
APERTURESyndicate OÜ
Republic of Estonia (EU) · Registry code 17384111 · VAT EE102972654
[email protected]

Publication & Review | AS Docs