Does ChatGPT Remember Past Conversations? (2026)

Yes. ChatGPT remembers across separate conversations, inside your own OpenAI account and only while memory is on. Two mechanisms do the work: saved memories, the details you explicitly ask it to keep, and reference chat history, which lets it draw on your past conversations. Both sit under Settings → Personalization, per the OpenAI Help Center, and each can be switched off on its own.

Two things regularly surprise people. Deleting a conversation does not delete the memory that came out of it — those are separate stores, cleared in separate places. And whether ChatGPT remembers you is a different setting from whether your conversations train OpenAI's models: the first is under Personalization, the second under Data Controls. This article sets out what ChatGPT retains today, where each control lives, and where its memory stops. Every vendor fact below was checked against OpenAI's own documentation on 2026-09-14.

Why ChatGPT brings up old conversations

If ChatGPT keeps surfacing something you mentioned weeks ago, the cause is usually one of two stores rather than a malfunction.

The first is a saved memory. Saved memories are not attached to the conversation that produced them; they are account-level details that any later chat can draw on, which is why something you mentioned once in an unrelated thread can reappear in a fresh one. Per the OpenAI Help Center, saved memories are "kept until you delete them."

The second is reference chat history. With that setting on, ChatGPT "will reference your past conversations to recall useful information you've shared," per the same documentation, and relevant information from those conversations "may be added to new ones." This layer is less predictable than saved memories: details drawn from chat history "can change over time as ChatGPT updates what's more helpful to remember."

So the practical question is which store is responsible. If the detail is something you asked ChatGPT to remember, look in your memory settings. If it is something it simply picked up, the chat-history setting is the lever — and the original conversation is still a source until you delete that too.

What ChatGPT actually remembers

According to OpenAI's official memory documentation, memory is governed by two distinct settings:

• Reference saved memories. These are "details you have explicitly asked ChatGPT to remember, like your name, favorite color, or dietary preferences." Saved memories persist until you delete them, and unless removed they are always available to future responses.
• Reference chat history. With this on, "ChatGPT can also use information from your past chats to make future conversations more helpful." OpenAI's example: if you once said you like Thai food, it may account for that the next time you ask what to have for lunch.

The two behave differently over time. Saved memories are stable, kept until you delete them. Details drawn from chat history are not guaranteed to persist. As OpenAI puts it, "details from past chats can change over time as ChatGPT updates what's more helpful to remember," and "ChatGPT doesn't remember every detail from past chats, so use saved memories for anything you want it to always keep in mind."

The chat-history capability expanded in 2025. In an April 10, 2025 update, OpenAI announced that memory "now references all your past conversations to deliver responses that feel more relevant and tailored to you." A June 3, 2025 update on the same page notes that free users received "a lightweight version of memory improvements that provides short term continuity across conversations," while "memory for Plus and Pro users provides a longer term understanding of the user." The depth of chat-history recall therefore depends on your plan, even though the underlying feature is broadly available.

One clarification is worth making early: memories are not tied to individual conversations. As OpenAI states, "ChatGPT's memories evolve with your interactions and aren't linked to specific conversations." This has a practical consequence covered below.

How to see and control it

Every control lives under your profile, and OpenAI documents the exact paths.

To turn saved memories or chat-history referencing on or off, click your profile icon, then go to Settings → Personalization, per the reference documentation. You can enable saved memories only, both settings, or neither. The controls are independent, so you can keep explicit saved memories while switching off the broader chat-history behavior, or vice versa.

To review or remove what ChatGPT has stored, open Settings → Personalization → Memory, per the Memory FAQ. What you see there is a memory summary, which OpenAI describes as updated automatically as you chat and stamped with when it last changed. You can edit it by typing the change into the box at the bottom of the summary, or by highlighting a specific line and correcting it. The same page notes the summary "will not include everything that ChatGPT remembers based on your chats" — if you want to know whether a particular detail is held, the documented method is to ask in chat.

The three-dot menu on that page offers Delete and turn off memory. Per the Memory FAQ, this "does not delete your past chats," and if you turn memory back on later "ChatGPT may create new memories from chats that remain in your chat history, including older chats." Clearing the summary is therefore not the same as clearing the source material.

One deletion subtlety warrants attention. Turning memory off does not erase what was already saved, and deleting a conversation does not delete the memory that came from it. OpenAI is explicit in the reference documentation: "Turning saved memory off won't delete anything that's already been remembered. And deleting a chat doesn't remove saved memory from that conversation." To fully remove something, "delete both the saved memory in Settings and the chat where you originally shared it."

The full list of places a detail can persist is longer than most people expect. Per the Memory FAQ, to fully delete something ChatGPT may know about you, "you'll need to delete every source where it appears, including past chats, archived chats, files, the memory summary, and disconnect any connected apps that may contain that information." Archived chats and uploaded files are easy to overlook.

What ChatGPT memory does not do

Memory has real boundaries, and knowing them prevents both false expectations and false worries.

It does not follow you to other AI apps on its own. Memory is a feature of ChatGPT tied to your OpenAI account, and OpenAI provides no route that pushes saved memories or chat-history insights into another company's assistant. What exists runs the other way: rival products have built importers that read an export you hand them. Google's import-memory page describes uploading a data export as a .zip under Chat History Import, and Anthropic's import documentation describes pasting memory exported from another provider — both checked on 2026-09-14. Leaving is therefore a manual migration you perform, not a transfer ChatGPT offers. Because memory sits inside your account, it does sync across your own devices when you sign in to the same account. The Data Controls FAQ confirms account-level settings apply "to your entire account. It doesn't matter which device you're using."

Temporary Chat deliberately remembers nothing. For a conversation you do not want retained, OpenAI provides Temporary Chat. Per the Data Controls FAQ, Temporary Chats "are deleted from our systems after 30 days," aren't used to train OpenAI's models, "may be reviewed only to monitor for abuse," and "don't get saved in your history and don't create memories." The Memory FAQ adds that Temporary Chats "do not use existing memories or create new memories," so an existing memory will not be drawn on either. It is the appropriate way to ask a one-off question without touching your memory state.

Business plans handle memory differently. For Team, Enterprise, and Edu accounts, OpenAI states in the memory announcement that "memories and any other information on your workspace are excluded from training our models," and that "Enterprise account owners can turn memory off for their organization at any time." Individual users in those workspaces still control how and when their memories are used. If you use ChatGPT through an employer, memory behavior may be set at the organization level rather than by you alone.

Memory vs. context window

The most common confusion is between memory and the context window, which are distinct.

The context window is the working span of text the model can see within a single conversation: your current messages plus the model's replies, up to a size limit. It is short-term and local. When a conversation grows long enough, earlier turns fall outside that window and the model can no longer see them directly. This is not memory in the feature sense; it is the model's immediate attention, and it resets when you start a new chat.

Memory is the persistent layer that survives across separate conversations. Saved memories and chat-history insights are pulled in as relevant context to new chats, which is why ChatGPT can recall your name or preferences in a brand-new conversation even though that conversation's context window started empty. In OpenAI's framing, "relevant information from your past conversations may be added to new ones," similar to how custom instructions work.

A concise way to hold the distinction: the context window is what ChatGPT can read in front of it right now, and memory is what it carries in from before. Losing track of something mid-conversation is usually a context-window limit, not a memory failure.

Privacy controls

Memory and model training are separate settings, and conflating them leads to the wrong action.

Whether ChatGPT remembers you is controlled under Settings → Personalization. Whether your conversations are used to improve OpenAI's models is controlled separately under Settings → Data Controls, via the "Improve the model for everyone" toggle documented in the Data Controls FAQ. Turning off training does not delete your memories or your chat history; as OpenAI notes, "your conversations will still appear in your chat history but won't be used to train ChatGPT." Conversely, turning off memory does not by itself change your training setting.

OpenAI notes in the memory announcement that "we may use content that you provide to ChatGPT, including memories, to improve our models for everyone," and that you can turn this off through Data Controls. The company also states it steers ChatGPT "away from proactively remembering sensitive information, like your health details, unless you explicitly ask it to." As noted above, content from Team and Enterprise plans is excluded from training.

If your goal is a private, no-trace exchange, the strongest single control is Temporary Chat, which combines no memory, no history, and no training in one mode.

FAQ

Does ChatGPT remember me between separate conversations? Yes, if memory is on. Saved memories and referenced chat history are carried into new conversations, so ChatGPT can recall your name, preferences, and prior context even in a fresh chat. If memory is off, each conversation starts blank.

How do I make ChatGPT forget something specific? Tell it to forget the detail in conversation, or open Settings → Personalization → Memory and correct the memory summary directly. To fully remove it, also delete the original chat where you shared it, since deleting one does not delete the other.

Does my ChatGPT memory carry over to other AI tools? Not by itself. ChatGPT has no feature that sends your memory to another assistant. You can move it by hand, because other vendors accept an import: per Google's import-memory page you upload a data export as a .zip, and per Anthropic's import documentation you paste memory exported from another provider. Both were checked on 2026-09-14, and both describe an import you carry out, not an automatic link between accounts.

Is memory the same as ChatGPT's context window? No. The context window is the short-term text the model can see within one conversation and it resets each chat. Memory is the persistent layer that survives across conversations. Forgetting something mid-chat is usually a context limit, not a memory setting.