Memory and reset: what your dot keeps and what it deletes

Your dot is built to remember, which is what makes it useful across days and channels. It also means that deleting what it knows is a blunt tool right now: there is no per-memory delete, only a full reset. This page lays out where your dot’s memories come from, what you can control today, and exactly what a reset removes and leaves behind.

Where your dot’s memories come from

So memory flows both ways, and from three places:

• From ChatGPT. Your ChatGPT memories and recent conversation context.
• From your conversations with your dot. These can also feed back into ChatGPT memory.
• From connected apps. Your dot can review connected information proactively and form memories from it. See connect apps.

In the mobile app, Customize → Memory in your dot’s profile shows what ChatGPT remembers, using your shared ChatGPT memory settings.

What you can and can’t control today

You want to…	What OpenAI documents
Stop ChatGPT memories flowing to your dot	Turn off Memory in ChatGPT. This “stops that sharing,” but “does not delete information that your dot has already received.”
See or edit a single dot memory	Not possible yet. “You currently cannot view, delete or directly modify individual dot memories.”
Stop your dot learning from an app	Disconnect the plugin. It stops new access, but it “does not delete information your dot has already built into its context.”
Delete your dot’s own memories	Reset (delete) your dot. It is the only documented way.
Know how long it keeps context	“For as long as you keep your dot.”

OpenAI also says your dot’s context “does not retain credentials, images, or screenshots.”

What reset deletes vs. what survives

This table puts OpenAI’s two help articles side by side. It is the part to read before you press Reset.

Item	Deleted by reset?	Source
Your dot itself	Yes	Getting started: reset “deletes your dot”
Its conversations	Yes	Getting started
Its saved memories and context, including what it learned from connected apps	Yes	Getting started and safety FAQ
Its scheduled tasks	Yes	Getting started
Files your dot created	No, stored separately	Safety FAQ
Codex threads your dot created	No, stored separately	Safety FAQ
ChatGPT conversations your dot created	No, stored separately	Safety FAQ
Your ChatGPT memories	No, they remain	Safety FAQ
Your custom rules	Not documented	Not covered
Plugin connections in ChatGPT	Not documented	Not covered

How to reset your dot

OpenAI’s help center documents these steps:

1. Open your dot’s profile.
2. Choose •••, then Reset.
3. Review what will be deleted.
4. Choose Reset to confirm, or Cancel to back out.
5. To start again, create a new dot in the ChatGPT desktop app or on desktop web. See create your dot.

Pause vs. reset

If you only want your dot to stop working for a while, pause it instead.

	Pause	Reset
How	Profile → ••• → Pause	Profile → ••• → Reset → review → Reset
Reversible	“Paused • Tap to resume”	Create a new dot on desktop
Memories	OpenAI doesn’t describe pause as deleting anything	Deleted
Scheduled tasks	OpenAI doesn’t describe pause as deleting anything	Deleted
Good for (our reading)	Travel, a busy week, or checking something before it acts again	Removing what your dot learned, or starting over

Before you reset: a checklist

• Copy anything you need from your conversations. They are deleted with your dot.
• Write down your scheduled tasks. You’ll have to recreate them. The scheduled tasks template makes that quicker.
• Note your custom rules. OpenAI doesn’t say whether they survive a reset. See custom rules.
• Disconnect apps you no longer want in use. Reset deletes what your dot learned, but plugin permissions are shared with ChatGPT, ChatGPT Work and Codex.
• Check ChatGPT memory separately. It is not touched by a reset.

Training and human review

OpenAI’s safety FAQ covers what happens to your dot’s data beyond memory:

• Personal plans. Whether your content is used to improve models is governed by the “Improve the model for everyone” setting.
• Business, Enterprise and Edu. Content is not used for training by default.
• Background research. “We don’t train directly on proactive background research or its notes.”
• Human review. It may occur in limited circumstances, even with model improvement turned off.
• Data requests. You can make a data subject access request through OpenAI’s Privacy Portal or by emailing dsar@openai.com.

For how dots fit into ChatGPT as a whole, see what ChatGPT dots are.