Telling a chatbot “forget that” almost never works, because the model has no power over what the app stores. To make an AI chatbot forget something, delete it where it lives: the saved-memory list, the chat history, and any profile or character fields that mention it. For cloud apps, also check how long the company keeps deleted data on its servers, and request deletion if it matters.
Why “please forget that” doesn’t work #
When you ask a character to forget something, the model will usually agree, because agreeing is the natural reply. But the model doesn’t control memory. The app does. If the detail was saved as a memory, it’s still in storage and will come back the next time it’s relevant. If it’s in the recent messages, the model can still see it in the very next reply.
The only reliable way to remove something is to delete it from the place it’s stored.
Where an AI chatbot keeps what you told it #
| Location | What’s there | How to clear it |
|---|---|---|
| Recent messages | The last several turns, word for word | Delete the message, or start a new chat |
| Chat summary | A condensed version of older messages | Delete the chat or thread |
| Saved memories | Facts extracted about you | Delete or edit the memory entry |
| Profile and character fields | Your name, persona, character notes | Edit the field |
| Company servers | Stored chats, logs, backups (cloud apps) | App settings, then a deletion request |
| Training data | Chats used to improve models, if allowed | Opt out; past training is hard to undo |
Many people clear their chat history and assume the job is done, but saved memories are often stored separately. On some services, deleting a conversation leaves everything learned from it in place.
How to make an AI chatbot forget something, step by step #
- Find the memory list. Look for “Memory”, “Memories”, “Journal” or similar in the app or the character’s profile.
- Delete or edit the specific fact. Editing is better if the fact is partly right, such as the wrong name for your sister.
- Delete or regenerate the message if the detail is in the recent chat. Otherwise the model can still see it for a while.
- Start a new chat if the detail is woven through the current one. A fresh thread clears the recent window and summary, though not saved memories.
- Check your profile and the character’s fields. Details you wrote there are included in every reply.
- Turn off memory or training in settings if you don’t want new details saved or used.
- Ask the company to delete your data for cloud apps. Privacy laws such as the EU’s GDPR and California’s CCPA give many users a right to request deletion, and most privacy policies explain how.
- Delete everything as a last resort: clear all data or delete the account.
What if the AI keeps bringing it back? #
If a deleted detail keeps coming up, one of these is usually the cause:
- It’s in more than one place. You deleted the memory, but it’s still in the chat summary or the character’s description.
- Another memory implies it. Deleting “has a sister named Jo” won’t help if “Jo visited last weekend” is still saved.
- The model is inferring it. If the recent chat hints at the detail, the model may reconstruct it without any memory at all.
- The app hides some memory. Some apps keep summaries or state you can’t see or edit. Starting over with a new character may be the only fix.
Our overview of how AI companion memory works explains each of these layers in more detail.
Making a companion forget in Xin #
Xin keeps its memory visible, which makes forgetting straightforward. Each companion has a memory journal listing every saved fact. Tap one and choose Forget to delete it, or edit it if it’s only partly right. Chat threads can be swiped to delete in the Chats tab, which also removes that thread’s summary. And Me, then Erase all data, wipes every companion, chat and memory at once.
Because Xin runs its model on your phone and never uploads chats, there’s no server copy to worry about. Two exceptions: if you used the in-app report button, the excerpt you reported was sent to the developer, and your phone’s own iCloud or Google backups may hold a copy of app data until they’re replaced.
Frequently asked questions #
Does deleting a chat delete the AI’s memory of it? #
Not always. Many apps store extracted memories separately from chat history, so a deleted chat can leave saved facts behind. Check the memory list after deleting a conversation and remove anything you don’t want kept.
Can AI companies see chats I’ve deleted? #
For cloud apps, deleted chats may stay on servers or in backups for a period set by the privacy policy, and some may be kept longer for safety or legal reasons. Read the retention section of the policy. On-device apps don’t have a server copy.
Can I make an AI chatbot forget me completely? #
For cloud apps, delete your account and submit a data deletion request, then keep the confirmation. Data already used to train a model is difficult to remove, which is why opting out of training early matters. For on-device apps, clearing app data or uninstalling removes everything stored locally. For more on what’s collected in the first place, see what data AI chat apps collect.
Will the AI notice that I deleted a memory? #
No. The model only sees what the app includes in the prompt. Once a memory is deleted, it simply isn’t there. The character may still infer something from the recent conversation, but it has no record that anything was removed.