ChatGPT Says Memory Full. Here's What That Means and What to Do
What the "saved memory full" message actually means, how to clean up the list without losing what matters, and how to keep your history if you ever move.
TL;DR: "Saved memory full" means the list of facts ChatGPT keeps about you has hit its cap. It won't save anything new until you clear space. Here's how to clean it up without losing the entries that matter.
Keep your memories somewhere you can see them →
What the message actually means
ChatGPT keeps two separate things. There's the conversation you're in, and there's a list of saved facts about you that carries across all conversations. The "saved memory full" notice is about the second one.
That list has a size cap. When it fills, ChatGPT stops adding new entries. It doesn't delete old ones to make room, and it doesn't tell you which conversations are being affected. It just quietly stops learning anything new about you while continuing to work normally in every other way.
So nothing is broken and nothing was lost. Your existing memories are intact, your chats still work, and the model still uses everything it saved before the ceiling. The only thing that stopped is new memory.
Worth knowing: the cap is on the memory list, not on your chat history, and it has nothing to do with the context window inside a single conversation. Those are different systems that fail in different ways.
How to clear it, in order
1. Look at the list first. Settings, then Personalization, then Manage memories. Read the whole thing before deleting anything. Most people are surprised by what's in there.
2. Delete the dead weight. In practice, a large share of a full memory list is junk. Entries about a one-off task from four months ago. Things you asked once and never again. Preferences for a project that ended. Duplicates of the same fact phrased three ways. Clear these first, and you'll usually free most of the space you need without touching anything you care about.
3. Consolidate the survivors. Six separate entries about your job can become one. Delete the six, then tell ChatGPT in a fresh message: "Remember that I work as a [role] at [place], my focus is [x], and I prefer [y]." One dense entry does the work of several thin ones and uses less of the cap.
4. Copy anything you don't want to lose. Before any big cleanup, select the whole memory list and paste it into a text file. It takes thirty seconds and it's the difference between a cleanup and a loss. This file is also what you'd use if you ever move to another AI.
5. Turn off automatic saving if you're a heavy user. In the same settings panel you can stop ChatGPT from deciding on its own what to remember, and save things deliberately instead. The list fills much more slowly and the contents stay relevant. The trade is that you have to remember to say "remember this."
What not to do
Don't clear all memories as a first move. The button is right there and it's tempting. It deletes everything with no undo, including the entries that took months to accumulate and are the reason ChatGPT feels tuned to you at all.
Don't assume deleting a chat deletes its memories. Chat deletion and memory deletion are separate. A fact saved from a conversation you deleted is still on the list.
Don't count on the cleanup lasting. With auto-save on, a heavy user refills the list in weeks. If you've hit it once, you'll hit it again, so the habit of consolidating beats a single deep clean.
The part that keeps happening
The reason this comes up repeatedly is structural. Memory is capped, invisible until you go looking for it, and managed through a settings panel rather than being part of how you actually use the thing. You only find out it's full when it tells you, and by then it's been full for a while.
There's a different way to handle it. On lookatmy.ai memory is a visible object rather than a hidden list. It grows a bonsai tree at /tree where each branch is a theme and each leaf is one memory in plain language. You browse it, search it, and trim what's wrong or stale, with the trim reversible if you cut something you wanted.
That changes the maintenance problem into something you can actually do. Instead of discovering a cap, you can see what's accumulating and prune as you go, the same way you'd tidy any other thing you own. You can also run more than one tree, so work context and personal context don't compete for the same space, though extra trees need Pro or higher.
The other structural difference is what happens when a model goes away. ChatGPT users who bonded with 4o found out that memory surviving doesn't help much if the voice changes. Here the memory and personality sit on your account rather than inside one model, and you choose from 350+ models underneath. Switch mid-conversation, or move off a deprecated model entirely, and the history stays.
Plans are Starter at $4.99, Pro at $9.99 and Max at $19.99, with a free tier to try it before deciding.
If you'd rather move the list than manage it
That text file from step 4 is all you need. Paste it into the memory import on lookatmy.ai, either during onboarding or from the tree, and your companion starts already knowing what ChatGPT knew about you. There's a longer walkthrough in the ChatGPT memory export guide.
Either way, copy the list before you clear it. That's the one step people skip and regret.
Read next
Why Your AI Forgets You, and What Actually Fixes It
Context windows and memory stores are two different systems that fail in different ways. Here's what's happening when your AI forgets, and the habits that fix it.
How to Write an AI Character Card That Doesn't Fall Apart at Message 50
Most AI character cards collapse once the small talk ends. Here's a card structure that holds up over months, the two fields people skip, and how to test one.
Leaving Chai? An Alternative Where Memory Isn't the $30 Tier
Chai keeps long-term memory in its top tier and capped chats without warning in 2026. Here's a Chai alternative with a memory tree you can see and 350+ models.