Metlivi Blog

What Not to Share With a Fictional Character Chatbot: A Practical Privacy Checklist

If you use a chatbot to improvise a scene or build a story with a fictional character, keep your real identity and other people’s private details out of the conversation. You can usually give the character enough to respond well with invented names, broad settings, and story-specific details. Before you upload anything or connect an account, check the particular app’s current privacy controls and decide what information you are comfortable sharing with that service. This guide covers what you type, attach, or authorize; it is not a technical audit of how an app handles data behind the scenes.

September 27, 20268 min readHome, Safety, Pets & Sustainable LivingBy Metlivi Editorial Team
Section 1

A quick check before you send

Pause over each message, image, recording, or permission and ask: **Does the character need this exact real-world detail to continue the story?** If not, leave it out or replace it with a fictional equivalent. A useful rule is to keep the emotional or narrative meaning while removing the real-world identifier.

For example, “My character is moving to a new town and starting over” gives a roleplay partner a clear direction. Your full name, current address, exact moving date, employer, and the names of people involved are unlikely to improve the scene. Privacy International’s [guide to chatbot settings and good practices](https://www.privacyinternational.org/guide-step/5676/chatgpt-settings-and-good-practices) recommends limiting personal information in AI chatbot use and reviewing controls such as shared chats, connected apps, data use, and deletion. Its steps describe ChatGPT specifically; menus and available settings vary by service.

Section 2

Keep these details out of ordinary roleplay

Use this checklist before typing or attaching personal material. These are practical limits for a fictional conversation, not a claim that every app collects or uses each item in the same way.

- **Exact identity:** Avoid your full name, usernames used elsewhere, account numbers, identity-document details, and combinations of facts that point directly to you. Use a made-up character name. If the story needs a personal quality or background, give only the broad trait that matters, such as “a university student” or “a new neighbor,” rather than a school, workplace, or identifying timeline.

- **Contact details:** Leave out email addresses, phone numbers, social handles, mailing addresses, and personal website links. A character does not need a way to contact you outside the chat. If a scene involves a message or meeting, invent the contact information or keep it generic: “send a note through the guild’s message board.”

- **Precise location and routine:** Avoid a home or work address, live location, regular route, named local venue, or a schedule that reveals when you are away. Replace a real place with a broad or fictional setting: “a coastal town,” “a crowded station,” or “a café on the edge of the kingdom.” Keep landmarks and timing invented when they could narrow down where you are or when you will be there.

- **Other people’s private facts:** Do not include a friend’s full name, contact details, private messages, personal history, or sensitive circumstances just because they appear in the plot. Change identifying details and share only what the scene needs. Instead of reproducing a real conversation, describe the fictional conflict in general terms: “two friends disagree about whether to keep a promise.”

- **Screenshots and documents:** Screenshots can show more than the text you intend to discuss: names, profile pictures, notifications, browser tabs, account details, or nearby messages may also be visible. Avoid uploading real conversations, documents, account pages, or images containing identifiers. If an image is genuinely necessary for a fictional scene, create a fictional example or crop and redact identifying details first, then inspect the result for anything left visible. Redaction should remove the information itself, not merely place a movable mark over it.

- **Voice recordings:** A recording can include your voice and incidental details, such as someone else speaking nearby. Skip voice input if text will do. If you choose to use it, avoid saying names, addresses, or other identifying details, and check whether the app offers a voice-specific setting or explanation before recording. Do not record another person for the scene without considering their privacy and the app’s rules.

- **Connected-app access:** A chatbot may offer to connect to other services or accounts. Do not grant access simply to make a roleplay more convenient. Check which service is being connected, what information the permission screen says it can access, and whether the feature is necessary. If the story does not need files, calendar entries, or other account information, leave the connection off.

**Exact identity:** Avoid your full legal name, usernames used elsewhere, account numbers, identity-document details, and combinations of facts that point directly to you. Use a made-up character name. If the story needs a personal quality or background, give only the broad trait that matters, such as “a university student” or “a new parent,” rather than a school, workplace, or identifying timeline.
**Contact details:** Leave out email addresses, phone numbers, social handles, mailing addresses, and personal website links. A character does not need a way to contact you outside the chat. If a scene involves a message or meeting, invent the contact information or keep it generic: “send a note through the guild’s message board.”
**Precise location and routine:** Avoid a home or work address, live location, regular route, named local venue, or a schedule that reveals when you are away. Replace a real place with a broad or fictional setting: “a coastal town,” “a crowded station,” or “a café on the edge of the kingdom.” Keep landmarks and timing invented when they could narrow down where you are or when you will be there.
**Other people’s private facts:** Do not include a friend’s full name, contact details, private messages, personal history, or sensitive circumstances just because they appear in the plot. Change identifying details and share only what the scene needs. Instead of reproducing a real conversation, describe the fictional conflict in general terms: “two friends disagree about whether to keep a promise.”
**Screenshots and documents:** Screenshots can show more than the text you intend to discuss: names, profile pictures, notifications, browser tabs, account details, or nearby messages may also be visible. Avoid uploading real conversations, documents, account pages, or images containing identifiers. If an image is genuinely necessary for a fictional scene, create a fictional example or crop and redact identifying details first, then inspect the result for anything left visible. Redaction should remove the information itself, not merely place a movable mark over it.
**Voice recordings:** A recording can include your voice and incidental details, such as someone else speaking nearby. Skip voice input if text will do. If you choose to use it, avoid saying names, addresses, or other identifying details, and check whether the app offers a voice-specific setting or explanation before recording. Do not record another person for the scene without considering their privacy and the app’s rules.
**Connected-app access:** A chatbot may offer to connect to other services or accounts. Do not grant access simply to make a roleplay more convenient. Check which service is being connected, what information the permission screen says it can access, and whether the feature is necessary. If the story does not need files, calendar entries, or other account information, leave the connection off.
Section 3

Replace personal details without flattening the story

Privacy-conscious writing does not have to be vague. Keep the detail that drives the scene—a choice, relationship dynamic, or atmospheric cue—and fictionalize the clues that identify real people or places.

Examples: Your actual workplace and a named colleague → “A new team member is unsure how to respond to a blunt coworker.” A recent event at a recognizable local place → “At a busy neighborhood market, two characters run into someone from their past.” A friend’s personal message → “One character feels left out after a friend cancels plans.” A real daily schedule → “The character has only a short window before the next train.”

These alternatives preserve what a character needs to respond to while removing details that are incidental to the fictional exchange. If a detail still feels identifying after you have changed the name, change the setting, timing, or surrounding facts too. Names alone are not a reliable substitute when a cluster of details still points to the same person.

Your actual workplace and a named colleague → “A new team member is unsure how to respond to a blunt coworker.”
A recent event at a recognizable local place → “At a busy neighborhood market, two characters run into someone from their past.”
A friend’s personal message → “One character wonders whether to invite a friend again after a canceled plan.”
A real daily schedule → “The character has only a short window before the next train.”
Section 4

Check the app’s current privacy controls

Settings change, and different services offer different features. Read the current privacy notice and in-app controls for the app you are using rather than assuming that a setting in one chatbot applies to another. Look for clear answers to these questions:

Sharing: Can conversations be shared by link, and where can you review or remove links you created?

Data use: What does the service say it does with submitted text, images, and audio? Read the current explanation rather than inferring that one switch controls every use.

Deletion: What does the service actually say happens when you delete a chat? A visible delete button alone does not establish a retention period.

Connections: Which other apps or accounts have you connected, what can each connection access, and how can you remove one you no longer need?

Media and account controls: Are there separate settings for uploads, recordings, personalization, and account access that apply to the way you use this service?

**Can conversations be shared, and how are shared links managed?** Privacy International notes that ChatGPT conversation links can be accessible to people who have the link and advises checking a conversation for personal or sensitive information before sharing it. Review any sharing controls before creating or circulating a link.
**Can chat content be used for model improvement or another purpose?** Find the current control and its description. Do not assume that turning off one use means the service stops every other form of processing or retention.
**What happens when you delete a conversation?** Check the app’s explanation of deletion and retention; do not infer a specific deletion timeline from the presence of a delete button.
**Which apps or accounts are connected?** Review connected services and remove access you no longer need, both in the chatbot’s settings and, where applicable, in the connected service. Privacy International’s ChatGPT-specific instructions recommend reviewing connected apps and removing unused connections.
**Are there controls for audio, uploads, personalization, or account security?** Check the categories that match what you use. Privacy International’s guide also discusses account security and data controls for ChatGPT, but its instructions may change and do not establish what another app offers.
Section 5

A simple send-or-rewrite routine

Before submitting a prompt or upload, make six quick checks:

1. **Identity:** Replace real names and account identifiers with invented ones.

2. **Reachability and whereabouts:** Remove contact details, exact locations, and schedules.

3. **Other people:** Generalize private facts and paraphrase rather than pasting their messages.

4. **Attachments and audio:** Inspect screenshots for surrounding details; avoid real recordings when text will work.

5. **Permissions:** Decline connected-app access unless the story feature needs it, and check what access is requested.

6. **Controls:** Review the app’s current sharing, data-use, deletion, and upload settings before relying on them.

If you already sent a detail, check whether the app lets you delete the conversation or manage a shared link, then review the service’s current help and privacy information to understand what those controls do. Avoid repeating the detail in follow-up messages. For future scenes, keep a small set of reusable fictional names, places, and backstory notes so you can add vivid detail without borrowing identifying facts from everyday life.

**Identity:** Replace real names and account identifiers with invented ones.
**Reachability and whereabouts:** Remove contact details, exact locations, and schedules.
**Other people:** Generalize private facts and paraphrase rather than pasting their messages.
**Attachments and audio:** Inspect screenshots for surrounding details; avoid real recordings when text will work.
**Permissions:** Decline connected-app access unless the story feature needs it, and check what access is requested.
**Controls:** Review the app’s current sharing, data-use, deletion, and upload settings before relying on them.
Related reading

Keep exploring this topic