Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more

ChatGPT can now use relevant information from earlier conversations, but it does not literally remember every word you have ever typed. OpenAI’s April 2025 update expanded personalization beyond the saved-memory list: ChatGPT may draw on useful context from your chat history when answering in a new conversation. You control this behavior in Settings > Personalization, and you can use Temporary Chat when a conversation should not affect personalization.

What ChatGPT remembers

ChatGPT has two related memory mechanisms:

Mechanism How it works What you can do
Saved memories Specific details ChatGPT stores for future conversations, such as a preference or recurring personal context. Review, edit, delete, or turn off saved-memory referencing.
Reference chat history ChatGPT can select relevant information from earlier chats even when that information is not listed as a separately saved memory. Turn chat-history referencing off; OpenAI says derived information is then scheduled for deletion from its systems within 30 days.

The system does not retain or consult every detail from every conversation. OpenAI’s Help Center says ChatGPT “does not retain every detail from every conversation” and selects information it considers relevant to a response. A reply can therefore reflect an old preference without being a verbatim search of your entire archive.

How the feature evolved

Saved memories arrived first

OpenAI introduced saved memories in April 2024. The original model centered on explicit requests such as asking ChatGPT to remember a fact.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Chat-history reference expanded personalization

In April 2025, OpenAI announced that ChatGPT could reference useful context outside the saved-memory list. OpenAI later described the background process that curates relevant context as an early version of “dreaming,” with a reviewable memory summary. Availability can differ by plan, region, platform, and managed workspace.

How to see or change what ChatGPT uses

  1. Open ChatGPT and select your account menu.
  2. Choose Settings.
  3. Open Personalization.
  4. Review the controls shown for your account, which may include Memory, Reference saved memories, Reference chat history, Memory summary, and Manage.

You can also ask ChatGPT what it remembers and request that a remembered detail be changed or removed. Exact labels and available controls depend on your account and rollout.

How to stop ChatGPT from using past chats

Turn off reference chat history

In Settings > Personalization, switch off Reference chat history when that option is available. OpenAI says information derived from past chats is scheduled for deletion from its systems within 30 days. Your original conversations remain in chat history until you delete them separately.

Turn off saved-memory referencing

Switch off Reference saved memories to stop ChatGPT from using saved memories for personalization. OpenAI states that turning off saved-memory referencing also turns off reference chat history. If you want existing details removed, use the available memory-management controls rather than relying only on the switch.

What’s actually slowing this PC down?

Pick the symptom - the matching free tool is one click away.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Delete both the source chat and the saved memory when necessary

A saved memory and the conversation that supplied it are separate. Deleting a chat alone does not automatically delete a separately saved memory. For a complete cleanup, delete the saved item in memory management and delete the relevant conversation from chat history.

Use Temporary Chat for a conversation that should not personalize future replies

Temporary Chat is the clearest built-in choice for a one-off conversation that should not affect personalization. OpenAI says Temporary Chats:

  • Do not appear in chat history.
  • Do not create or update memories, even when personalization is enabled.
  • Are not used to improve models while they remain temporary.

Limited safety-related context can still apply in rare, high-risk situations, so Temporary Chat is not a promise that no safety system can process any signal.

Memory is different from model training

Personalization controls and model-training controls address different uses of your data.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
  • Memory settings govern whether ChatGPT can use saved memories or relevant information from prior chats to tailor replies.
  • “Improve the model for everyone”, found under Settings > Data Controls for eligible personal accounts, governs whether chats and remembered information may contribute to model improvement. Turning this setting off does not by itself erase saved memories or delete chats.
  • Business, Enterprise, Edu, and Healthcare workspaces can have different defaults and administrator-managed policies. Follow your organization’s workspace settings and privacy notice.

OpenAI describes memory as optional and says users can review, edit, delete, or turn it off. Its consumer privacy guidance also advises against sharing sensitive information that you would not want used or reviewed.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

What someone with account access might learn

If another person can access your unlocked account or device, they may be able to ask ChatGPT about information present in old chats or available memory. That is a privacy risk arising from account or device access, not a guarantee that ChatGPT will reveal every historical detail. Protect the account, use device authentication, and remove sensitive memories and chats you no longer need.

Which option fits your goal?

Your goal Best action Important limitation
Keep useful preferences across chats Leave the relevant personalization controls enabled. ChatGPT still selects what it considers relevant; it does not use every detail.
Stop future personalization from old chats Turn off Reference chat history. Derived information is scheduled for deletion within 30 days; original chats remain until separately deleted.
Remove a known personal detail Delete it under memory management and delete source chats if needed. Deleting only the source chat may leave a separately saved memory.
Have a private, non-personalizing one-off conversation Start a Temporary Chat. Rare safety-related processing can still apply.
Stop chats being used for model improvement on a personal account Turn off Improve the model for everyone in Settings > Data Controls. This is separate from deleting memories or chats.

The Bottom Line

ChatGPT remembers more across conversations, but not everything you have ever said. Check Settings > Personalization, delete saved memories and source chats separately when required, and use Temporary Chat for conversations that should not update personalization.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.

Free tools Windows power users keep installed

One-click scans. No signup required.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.