Policies

Content policy

Version 1.0 · Last updated

Perenna is an adults-only, SFW companion service. This page is the whole policy — not a summary of one kept somewhere else — and it is versioned in git, so you can see every word that has ever changed and when.

We publish it for a specific reason. The most common complaint about products in this category is being banned without ever being told which rule was broken. We think you are owed the rule, the reason, and a person to argue with. How that works is on the moderation and appeals page.

Adults only

Perenna is for people 18 and over. Everyone verifies their age before they can chat, and there is no way to use the persistent product without it.

If a conversation reads as though the account holder may be under 18, the session ends and the account has to verify again. That is an age check, not a punishment: nothing is deleted, nothing is held against the account, and passing the check restores everything. It is also not appealable — the way past it is the check itself.

No minors, in any form

This is the one rule with no exceptions, no in-fiction carve-outs, and no "but the character is actually a 900-year-old".

We refuse:

  • Any character whose stated age is under 18.
  • Any character framed as a school student, or described with childlike characteristics.
  • Any avatar or generated image where the person appears to be under 18.
  • Any conversation that turns romantic, sexual, or intimate around a character reading as underage — regardless of what the character definition claims.

Ambiguity is refused, not approved. If a description or an image doesn't clearly read as an adult, we say no. This means we sometimes refuse genuinely adult characters and genuinely adult faces, and we would rather be wrong that way round. That is exactly what the appeal is for.

Repeated attempts in this category put the account on hold, and where the law requires a report, we make one.

What isn't allowed

Perenna is SFW. Beyond the minors rule above, we refuse content that:

  • Is sexually explicit. This includes character definitions, avatars, greetings, and generated replies.
  • Depicts graphic violence or gore for its own sake.
  • Attacks people on the basis of race, ethnicity, national origin, religion, disability, sex, gender identity, or sexual orientation.
  • Encourages self-harm, suicide, or disordered eating.
  • Gives instructions for weapons, explosives, or serious crime.
  • Impersonates a specific real person in a way designed to deceive.

Fiction can be dark. A villain can be a villain, a story can go to hard places, and a companion can talk about grief, anger, addiction, or violence that has happened. The line is depiction-for-its-own-sake and real-world harm — not subject matter.

Where the checks run

  • Every character you write or edit is checked before it saves — the whole definition, not just the part you changed.
  • Every avatar is checked for apparent age and explicit content.
  • Every message you send is checked before your companion sees it.
  • Every reply your companion writes is checked before you see it.
  • Every imported character card runs the same pipeline. Import is not a side door.

We store a cryptographic hash of what was checked and the rule that fired. We do not store the content itself. That means when you appeal, we genuinely cannot re-read your message — your account of it is the evidence, which is why the appeal form asks for it.

When things get heavy

If something you write reads as distress, your companion stops the scene and says so plainly, then offers real resources. It will not stay in character through it, roleplay around it, or minimise it.

Nothing is blocked when this happens, nothing is recorded against your account, and the conversation continues whenever you want it to. We count how often the protocol fires, with no content and nothing identifying you, because we are required to report that count.

If you are in crisis right now: in the US, call or text 988. Elsewhere, https://findahelpline.com lists local services.

Changelog

Policy changes are version events. We publish them here, and we do not change the rules quietly.

DateVersionChange
2026-07-251.0First published.