How to Set Up a Private CrushOn.AI World Card

1. Create a world with a narrow fictional purpose

Expected result: an editor for your own small setting, with a clear reason to separate its lore.

Documented route: home bottom-right + > Create World > Manually Create > Name Your World.

Suggested name: 'Lantern Archive': adult curators investigate a misplaced fictional token. Avoid real-life identities. If creation is absent, stop; do not assume subscribing will reveal it.

The character-versus-world explanation separates a character's personality from shared setting information. Put the curator's dry humor in their character setup; use the world for archive facts. Ask whether a second fictional curator would need the same information. If not, a simpler character setup may suffice.

2. Add one recognizable keyword note

Expected result: one saved entry with a specific activation phrase and an easily checked fictional detail.

Edit Notes > Items > +. Fill Note name, Description, Trigger type: Keyword, Keywords and Star rating; save.

Suggested note name: 'Archive token'. Description: 'The brass token bears a seven-point star. Adult curators use it to request a catalog drawer; it is not a key.' Suggested keyword: 'brass token'. Start with a moderate rating rather than marking every detail as essential.

A distinctive engraving gives you something to check without supplying the answer in your prompt. Add neighboring artifacts only after this entry works.

3. Review basics and save as Private

Expected result: you have deliberately selected the intended audience before using the card.

World Info contains World Name, Cover Image, Intro, Tags, Rating and Visibility. Save And Publish > Private. Private: only you attach; Unlisted: link-accessible, not searchable; Public: discoverable.

Suggested intro: 'Adult curators investigate a fictional archive mystery.' Choose an accurate rating; this rehearsal needs no explicit material. Review cover rights and identifying details. Check the saved visibility.

Do not substitute Unlisted for personal-only use or circulate a link as a privacy experiment. Review character visibility separately; one asset's audience does not establish another's.

4. Attach and inspect a positive trigger

Expected result: the active-note list identifies your entry, and the reply can be checked against its contents.

Chat: + > World > select your card. Open the active-notes chip beneath the reply. If hidden, check Show Notes Bubbles in the management panel.

Use an appropriate original adult character and a disposable fictional conversation. Suggested prompt: 'The curator examines the brass token. Describe its marking and intended use.' Inspect the named entry, then compare the reply with your saved description: seven-point star, catalog request, not a key.

Record activation and accuracy separately. A pleasing answer without activation is inconclusive; an active entry with an invented engraving needs correction. If attachment is unavailable, resolve eligibility before spending more messages.

5. Check accidental matches and missed phrases

Expected result: a small record distinguishes deliberate hits, unexpected hits and unclear evidence.

Matching uses substrings; multi-word phrases require exact wording. Case and punctuation are ignored; full-width characters differ.

Avoid a short keyword such as 'ash': it also occurs inside 'cash'. For your own invented location, prefer a distinctive phrase such as 'Ash Observatory'. Add a separate alias only when you expect to use it, not a list of common words that could occur anywhere.

Suggested neutral prompt: 'The curator describes today's weather.' Try a separate fresh chat with the same card and inspect activation. No artifact entry is the intended outcome, not a guarantee. Next try an unconfigured alternate name such as 'catalog token'; expected result: the artifact entry does not activate, subject to the same uncertainty. These comparisons cannot establish undocumented scan depth or history-clearing rules.

Mistakes, limits and confidentiality checks

Keep the first card small and treat surprising replies as a reason to inspect, not a reason to reveal more information.

Runtime capacity can drop whole notes, prioritizing higher stars. Avoid adding a large archive before your basic check is understandable. Change one thing at a time and repeat the relevant prompt, so you know which revision you are evaluating.

Private visibility does not establish encryption, provider exclusion, training restrictions or deletion guarantees. The checked setup documentation is not a confidentiality promise. Review the current privacy policy before sharing anything sensitive; if your decision depends on an unanswered retention or access question, keep that information outside the service.

Exclude passwords, contact details, workplace material and others' private stories. Fictional names are insufficient if events identify someone. Record invented test phrases and outcomes, without account screenshots.

Stop, choose an alternative, or continue

Stop if access, audience or spending remains uncertain; continue only with a bounded fictional setup.

Pause after repeated unclear activation rather than buying a different tier on speculation. If you only need one short scene, keep a harmless fictional premise in the character conversation where your account already permits it. For material requiring confidentiality, use an appropriately protected local document instead of uploading it.

Summary: confirm account availability, save a small personal-only world, inspect a deliberate trigger and compare a neutral message. Keep a record of what you actually observed. Expand with one additional fictional entry only after those checks make sense; there is no need to change the audience to finish this workflow.

Frequently asked questions