Check What Your AI Chatbot Remembers About You, and Make It Forget

The tip

The big AI assistants no longer start every conversation from zero. ChatGPT, Google Gemini and Claude can all remember things you have told them and use them in later answers: your job, your family, a trip you are planning, a health worry you mentioned once.

That can be handy. It can also surprise you when an old detail turns up in a new answer, or when you realise you shared something you would rather the assistant did not keep. Set aside fifteen minutes and do three things in each assistant you use:

1. Look at what it remembers about you.
2. Delete or correct anything you do not want it to keep.
3. Learn the private mode (temporary or incognito chat) for the next time you want to ask something without it being remembered.

The steps below come from each company's own help pages, checked on the day this tip was published. Menus differ by plan, country, platform and app version, and all three companies say some controls may not appear for every account. If a label on your screen is slightly different, look for the closest match.

Why it matters

Memory changes the deal you make when you type something into a chatbot. Without memory, a detail you mention stays in that one conversation. With memory, it can follow you into every future conversation.

The companies are open about this. OpenAI says that when Memory is on, ChatGPT can remember relevant details "from your chats and other available sources," which depending on plan and region can include past chats, saved memories, custom instructions, saved files and content from connected apps such as Gmail. OpenAI also warns that "sensitive information can appear in Memory if you share it with ChatGPT." Google says that with memory of past chats turned on, Gemini "can learn from your chats to understand more about you and your world." Anthropic says memory is on by default for Claude's Free, Pro, Max and Ultra plans.

The most useful thing to understand is that deleting a chat and deleting a memory are often two different jobs. People assume that removing a conversation makes the assistant forget it. As you will see below, that is not always true.

ChatGPT: see, fix and delete what it remembers

Find the memory settings

OpenAI gives this path:

1. Open Settings.
2. Select Personalization.
3. Select Memory.

Depending on your account, OpenAI says you may see controls called Memory, Reference saved memories, Reference chat history, Memory summary, Manage or Saved memories. Not everyone sees all of them.

You can also simply ask ChatGPT what it remembers about you. OpenAI lists this as a way to review your memory.

What each control does

• Memory is the main switch. OpenAI says turning it off does not delete your past chats, and if you turn it back on later, ChatGPT may create new remembered information from chats still in your history.
• Reference saved memories lets ChatGPT use specific details it has saved. OpenAI says turning this off also turns off Reference chat history.
• Reference chat history lets ChatGPT use relevant information from past conversations. OpenAI says that if you turn it off, information remembered from past chats is scheduled for deletion from OpenAI systems within 30 days, while the original chats stay in your history unless you delete them. Turning it off does not delete your saved memories.
• Saved memories are details you asked ChatGPT to remember, or that it saved as useful context. OpenAI says these are stored separately from chat history.
• Memory summary, where available, is a high level view of what ChatGPT may use to personalize answers. OpenAI says it does not necessarily include every detail or source.

Correct, hide or delete something

From the memory summary, OpenAI says you can type a correction, highlight text and correct it, choose Don't mention this again where offered, or choose Delete and turn off memory from the more options menu (the three dots).

Know the difference between these:

• Don't mention this again reduces future references to the information. OpenAI says it does not delete the original source.
• Delete and turn off memory deletes the remembered information in the summary and turns Memory off. It does not delete past chats.
• Deleting a saved memory stops it being used for future personalization, but OpenAI says it does not remove mentions of that information from past conversations.

To truly remove something, clean every place it lives

OpenAI is clear that "deleting a chat alone does not necessarily delete a separate saved memory created from that chat." To remove a piece of information properly, OpenAI says to delete the saved memory and the chat where you first shared it, and also to remove it from:

• the memory summary
• saved memories
• regular and archived chats
• saved files in Library or Space
• connected apps (disconnect the app if you no longer want ChatGPT to access it)

OpenAI adds that changes can take time to take effect, and that it may keep logs of deleted saved memories for up to 30 days for safety and debugging.

Check where an answer came from

When ChatGPT uses your personal context, OpenAI says Sources may appear below the response. Opening it can show a past chat, saved memory, custom instruction, file or email that shaped the answer. Depending on the source, you may be able to correct or delete a saved memory, delete a referenced chat, or mark a source as relevant or not relevant. OpenAI notes that Sources may not show every factor.

Use a temporary chat for one off questions

OpenAI's steps:

1. Open a new chat.
2. Select Temporary.
3. Choose Personalized or Unpersonalized before you send your first message.
4. Enter your message.

What it does, according to OpenAI:

• A temporary chat stays out of your chat history and is not used to improve OpenAI models while it stays temporary.
• It does not create or update memories, even if you chose Personalized.
• Unpersonalized does not use your memory, custom instructions or plugins. Personalized can use your existing memories. You cannot change this after the chat starts.

What it does not do:

• It is not erased instantly. OpenAI says it may keep a copy for up to 30 days for safety purposes.
• If you save it, it becomes a regular chat and follows your normal memory and model improvement settings.

Google Gemini: memory of past chats

Who has it

Google says personalization from memory of your past chats needs you to be 18 or over, signed in with a personal Google Account (not work, school or supervised accounts), and to have Keep Activity turned on. Google lists it in the Gemini mobile app, the Gemini web app, Gemini in Chrome where that is available, and Gemini on your smartwatch, but not in some features such as Gems or Live chats.

If Keep Activity is off, this memory feature is not available. Our earlier tip explains Keep Activity and model training: Check Whether Your AI Chats Are Used to Train the AI.

Turn Memory on or off (computer)

1. Go to gemini.google.com.
2. At the bottom, click Settings, then Personal Intelligence.
3. Turn Memory off or on.

Google's help page has separate tabs for Android and iPhone or iPad, so if you use the app, follow the version for your device.

Check whether Gemini used a past chat

Google suggests simply asking Gemini: "Did you use any info from past chats?"

Delete or correct what Gemini knows

Google's approach is different from ChatGPT's. There is no separate list of saved facts described on this page. Instead:

• To delete something, Google says to delete all chats that contain that information from Gemini Apps Activity. There might be a short delay before Gemini stops using a deleted chat.
• To correct something, make sure Memory is on, then correct Gemini directly in a chat. Google adds that Gemini "might not always get it right."

Information from connected apps needs two steps

If Gemini picked something up from a connected Google app, such as Gmail or Google Photos, Google says you must do both:

1. Delete all chats containing that information from Gemini Apps Activity or Recent chats.
2. Disconnect the app that holds it.

Google explains why: if you only disconnect the app, Gemini might still use the information from a past chat. If you only delete the chats, Gemini might still find it in the connected app. Google also says that when you delete or change data in a connected app, Gemini may not reflect it until days later.

Use a temporary chat

On a phone, Google's steps are:

1. Go to gemini.google.com or open the Gemini app.
2. At the top, tap Menu.
3. Next to New chat, tap Temporary chat.
4. Type your question and submit it.

Google says temporary chats need a personal Google Account, so they are not available on work or school accounts, and they are not available in Gemini in Chrome or Gemini in Google Messages. In a temporary chat you cannot use Gems or Connected Apps, get personalized responses, save information for future personalization, or give feedback.

On data, Google's Privacy Hub says temporary chats are retained with your account for 72 hours, are not used to train Google's AI models, and are still used to respond to you and to help protect Google, its users and the public, including with help from human reviewers. When you leave a temporary chat, you can no longer open it.

Claude: memory topics, incognito and per chat switches

See everything Claude remembers

Anthropic says you can see what Claude remembers in Settings, then Your Claude, by selecting Manage next to Your memory. In the older settings layout it is Settings, then Memory. Everything is listed under Topics. Select a topic to read it, then use the edit icon to change it or Delete to remove it.

You can also tell Claude in a chat what to remember, change or forget, and Anthropic says the update applies to your next conversation.

What Claude does and does not save

Anthropic says Claude remembers everyday context such as your role, projects, the people and places in your work and life, and your preferences. By default, it says Claude does not store sensitive topics such as health, race, ethnicity, religious beliefs, politics and gender identity unless you turn on Include sensitive topics in memory. Anthropic also says some information is never saved even if you ask, including government ID numbers, criminal history, financial account numbers and immigration status.

Turn memory off for your whole account

In Settings, then Your Claude, the setting is Generate memory from chats. When you switch it off, Anthropic says you get two choices:

• Pause memory: Claude keeps what it already remembers but will not use it or make new memories. Chats while paused are not added later if you turn it back on.
• Reset memory: permanently deletes all memories, including project memories. Anthropic says this cannot be undone.

A separate switch, Search and reference chats, controls whether Claude can search your past conversations. Anthropic says past chat search is available on paid plans.

Turn memory off for one chat

Anthropic says you can keep memory on for your account and switch it off for a single chat, on the web, Claude Desktop and the latest Claude mobile app. You must do it before your first message:

1. Start a new chat.
2. Click or tap + in the message box.
3. Turn off Memory.
4. Send your first message.

Claude will not use your memory in that chat and nothing from it is saved to memory. Unlike incognito, the chat stays in your history.

Use an incognito chat

When you start a new chat outside a project, Anthropic says you will see a ghost icon in the upper right corner. Clicking it opens an incognito chat that is not saved to your chat history or to Claude's memory, and Claude will not pull from it when searching past conversations. Anthropic notes that on Team and Enterprise plans, incognito chats are still included in the organization's data exports and follow its retention policies.

Deleting a chat does not delete the memory

Anthropic says that when a conversation is deleted, "memory entries generated from it aren't removed," but you can delete individual memories while memory is on. So if you regret sharing something, delete the chat and check your memory topics.

What memory controls do not do

• They are not the same as model training settings. Memory is about personalizing your answers. Whether your chats help train the company's AI is a separate switch in each app. OpenAI, for example, puts that under Data controls as Improve the model for everyone.
• Private modes are not instant deletion. OpenAI says a temporary chat copy may be kept up to 30 days. Google says 72 hours. Read each company's wording before treating a chat as gone.
• They do not cover work accounts in the same way. On work and school plans, administrators often control memory. OpenAI and Anthropic both say workspace or organization owners can manage memory settings, and Google's memory feature is not available on work or school accounts at all.
• Safety uses continue. OpenAI says turning off Memory does not disable limited safety uses of context in rare, high risk situations. Google says even temporary chats are used to help protect users, including with human reviewers.

Common mistakes

• Deleting the chat and stopping there. In both ChatGPT and Claude, a memory made from a chat can outlive the chat. Check the memory list too.
• Choosing "Don't mention this again" when you meant delete. In ChatGPT that option changes behaviour but keeps the source.
• Forgetting connected apps. If an assistant can read your email or photos, deleting chats may not be enough. Google says to disconnect the app as well.
• Saving a temporary chat by accident. In ChatGPT, saving turns it into a normal chat that follows your memory settings.
• Assuming one assistant's settings carry over. Each app has its own memory. Clean each one you use.
• Never looking again. Memory keeps growing as you chat. A quick look every few months keeps it accurate.

Related checks worth doing now

1. Model training. While you are in settings, decide whether your chats can be used to improve the company's models. See Check Whether Your AI Chats Are Used to Train the AI.
2. Connected apps. In each assistant, review which apps (email, photos, files) it can reach, and disconnect any you do not use.
3. Family accounts. If your teenager uses ChatGPT, OpenAI says a parent can manage supported memory controls through parental controls. Our guide: Set Up Parental Controls for the AI Chatbots Your Child Uses.
4. Shared devices. If someone else uses your computer or phone, remember that a signed in assistant may show your chat history and personalized answers to them.

A worked example

Here is how a clean up might look for an imaginary user. The person is made up; the steps are the ones the companies describe.

Dana uses ChatGPT at home and Claude for work writing. A few months ago she told ChatGPT about a job interview she did not want her current employer to know about, and she mentioned a family medical appointment to Claude. Today she wants both gone.

ChatGPT. Dana asks ChatGPT, "What do you remember about me?" The answer mentions the job search. She opens Settings, then Personalization, then Memory, and finds the job search in her saved memories, so she deletes it. Because deleting a memory does not remove it from past chats, she also deletes the chat where she first mentioned the interview, and checks her archived chats. She has no connected apps, so there is nothing to disconnect. She leaves Memory on, because she likes ChatGPT remembering her cooking preferences.

Claude. Dana opens Settings, then Your Claude, and selects Manage next to Your memory. She reads through her topics. She has never turned on sensitive topics, but she checks anyway and finds nothing about the appointment. She deletes the old chat that mentioned it. Because Anthropic says deleting a chat does not remove memory entries made from it, she is glad she checked the topics first.

Next time. For sensitive one off questions, Dana now starts a ChatGPT temporary chat set to Unpersonalized, or a Claude incognito chat using the ghost icon, before she types anything.

Total time: about fifteen minutes.

The simple rule underneath it all

Memory makes AI assistants more useful, and you do not have to turn it off to stay safe. What matters is knowing it is there, looking at it now and then, and using a temporary or incognito chat before you type something you would not want an assistant to bring up again later.

Frequently asked questions

If I delete a chat, does the AI forget what I said in it?

Not always. OpenAI says deleting a ChatGPT chat does not automatically delete a saved memory made from it, and Anthropic says deleting a Claude conversation does not remove memory entries generated from it. For Gemini, Google says to delete the chats that contain the information.

Is a temporary or incognito chat completely private?

No. It stays out of your history and memory, but the company can still keep a copy for a while. OpenAI says up to 30 days for safety purposes, and Google says 72 hours. Anthropic says incognito chats on Team and Enterprise plans follow the organization's data retention policies.

Is memory the same as letting the company train its AI on my chats?

No. Memory personalizes answers for you. Model training is a separate setting in each app. Turning one off does not turn off the other, so check both.

Sources

1. OpenAI Help Center: Memory in ChatGPT
2. OpenAI Help Center: Temporary chat in ChatGPT
3. Gemini Apps Help: Get personalization with memory of your past Gemini chats
4. Gemini Apps Help: Use Gemini Apps (temporary chats)
5. Gemini Apps Help: Gemini Apps Privacy Hub
6. Claude Help Center: Use Claude's chat search and memory to build on previous context
7. Claude Help Center: Use incognito chats