How to Delete What ChatGPT, Gemini, and Claude Remember About You
AI chatbots are getting better at remembering you. ChatGPT, Gemini, and Claude can store details about your preferences, work, projects, writing style, and even people you mention often. That can make conversations feel more personal and useful, but it also raises an important privacy question: how do you delete what an AI chatbot remembers?
The answer depends on which chatbot you use. Deleting a conversation is not always enough. In some cases, the chatbot’s memory is stored separately from your chat history, which means personal details may remain even after the original conversation is gone.
Here is how memory works in ChatGPT, Gemini, and Claude, where to find the settings, and how to stop each AI tool from saving information you do not want it to keep.
ChatGPT memory is separate from your chat history
ChatGPT can remember details from previous conversations to personalize future replies. These memories may include your preferred writing tone, ongoing projects, profession, hobbies, or other information you have shared.
You can find ChatGPT’s memory controls by opening Settings, then going to Personalization and Memory. Depending on your account and region, you may see either a memory summary or a list of saved memories.
The important thing to know is that ChatGPT memories are not the same as chat history. OpenAI stores saved memories separately, so deleting a chat does not automatically delete a memory created from that chat.
To fully remove something from ChatGPT, you may need to delete it from multiple places, including:
The memory summary or saved memories section
The original chat where the detail appeared
Archived conversations
Uploaded files
Connected apps, if they were used
Custom instructions, if the information is stored there
If you tell ChatGPT “don’t mention this again,” that does not necessarily delete the information. It may reduce future references, but the original source can still remain unless you remove it directly.
For private one-off conversations, ChatGPT offers Temporary Chat. Temporary chats do not create new memories, but they may still use existing memories unless you choose an unpersonalized option before starting. Once a temporary chat is saved, it becomes a regular chat and may once again contribute to personalization.
Claude stores memory as topics
Claude’s memory system works differently. Instead of only summarizing old conversations, Claude can save individual topics while you chat. These topics might include your job role, current projects, important people, writing preferences, or recurring tasks.
Claude memory is enabled by default on several plans. You can find it by going to Settings and Your Claude, or by opening the Memory section in older settings layouts. From there, choose Manage to view the topics Claude has saved. You can open each topic, edit it, or delete it.
The key privacy detail is this: deleting a Claude conversation does not delete the memory topics created from that conversation. If Claude learned something from a chat, you must remove that memory topic separately.
If you want to wipe everything at once, you can turn memory off and choose Reset memory. This deletes saved memory and project memories, but it cannot be undone. If you only want Claude to stop using memory temporarily, use Pause memory instead. Pausing memory stops Claude from using and adding to memory, but it does not delete what has already been saved.
Claude also has some limits on what it saves by default. Sensitive topics such as health, religion, political views, and ethnicity are not saved unless you enable sensitive topic memory. Some information, such as government ID numbers, financial account numbers, criminal history, and immigration status, is not saved even if requested.
For conversations you do not want saved to memory, Claude offers two useful options. You can open an incognito chat using the ghost icon, which keeps the conversation out of chat history and memory. You can also turn off Memory in the plus menu before sending your first message. That keeps the chat in your history but prevents Claude from saving anything from it.
Gemini memory is tied to saved chats
Gemini handles memory in a more chat-based way. Its personalization comes from saved conversations, so remembered details are generally connected to your Gemini chat history.
To use Gemini personalization, you need a personal Google Account, you must be at least 18, and activity saving must be enabled. On the web, the setting is found under Settings and Personal context.
In many regions, Gemini does not yet offer a detailed list where you can delete individual remembered items. Instead, you need to delete the chats that contain the information you want removed. This is done through Gemini Apps Activity. Google notes that there may be a short delay before Gemini stops using deleted information.
If Gemini has access to connected apps, such as Gmail or Google Photos, deleting chats may not be enough. You may also need to disconnect the relevant app. Otherwise, Gemini could rediscover the same information later. On the other hand, disconnecting an app without deleting old chats may still leave Gemini with information it previously gathered from conversations.
In the United States, Gemini has more advanced memory controls. Some users can ask Gemini to remember specific details, and memory summaries have started appearing for certain accounts. These summaries allow users to view remembered information by category and delete it more directly. Asking Gemini to remember details may require a Google AI plan and availability can depend on location.
For questions you do not want saved, Gemini offers temporary chat. Temporary chats do not appear in activity, do not save information for future personalization, and do not provide personalized responses based on past chats.
Deleting a chat may not delete the memory
The biggest mistake many users make is assuming that deleting a chat erases everything an AI learned from it. That is not always true.
In ChatGPT, saved memories are separate from conversations. You must delete the memory itself.
In Claude, memory topics remain even if the original chat is deleted. You must remove the topic manually or reset memory.
In Gemini, remembered information is more closely tied to saved chats, so deleting the relevant conversation is usually the main step.
This difference matters if you have shared sensitive information, such as medical details, personal relationships, workplace issues, financial concerns, or private plans. Even casual comments can become part of a chatbot’s personalization system.
How to check what your AI chatbot remembers
A quick way to start is simply to ask.
In ChatGPT, you can ask: “What do you remember about me?”
In Gemini, you can ask whether it used information from previous chats to answer you.
In Claude, the best method is to open the memory management page and review the saved topics directly.
If you see something outdated, incorrect, or too personal, delete it from the right place. For ChatGPT and Claude, that means removing the memory entry or topic. For Gemini, that usually means deleting the conversation where the detail appeared.
Do not confuse memory with AI training
Memory and model training are not the same thing. Turning off memory may stop a chatbot from personalizing replies based on saved details, but it does not always mean your conversations are excluded from being used to improve AI systems.
If privacy is important to you, check both settings:
Memory or personalization settings
Data sharing or model training settings
These controls are usually separate. For the strongest privacy, disable memory, review training preferences, avoid connected apps when unnecessary, and use temporary or incognito chats for sensitive questions.
Best privacy habits for using AI chatbots
If you use ChatGPT, Gemini, or Claude often, it is worth reviewing memory settings regularly. AI memory can be helpful, but it should not become a hidden archive of information you forgot you shared.
Use temporary or incognito chats for sensitive topics.
Review saved memories or topics every few weeks.
Delete outdated or personal details manually.
Be careful with connected apps like email, photos, calendars, and files.
Turn off memory if you prefer a clean slate.
Check model training settings separately from memory settings.
Avoid sharing private identification, financial, or medical details unless absolutely necessary.
AI assistants are becoming more personal, but that also means users need to take a more active role in managing what these tools remember. A few minutes in the settings can prevent old conversations, sensitive details, or personal context from following you into future chats.
