ThreadRecall
How to

ChatGPT says memory is full — what it means and the real fix

The message sounds alarming and the fix sounds tedious. Both impressions are wrong. Here is what is actually full, how to clear it in two minutes, and the bigger problem the message quietly points at.

Tightly wound coil of teal thread filled to capacity, for the guide to ChatGPT's memory full message

At some point ChatGPT tells you its memory is full: a small notice that new memories cannot be saved until you make room. Most people react one of two ways. Some panic slightly, assuming their conversations are about to be deleted. Others sigh, assuming they now have to do digital housekeeping forever.

Neither is quite right. The message is about one small, specific thing, and clearing it takes a couple of minutes. But it is also a useful moment to understand what ChatGPT's memory actually stores, because it is much less than most people assume.

What "memory full" actually means

ChatGPT's memory has historically worked in two separate ways, and the distinction is the whole story here. OpenAI describes both in its Memory FAQ:

  • Saved memories are short, distilled facts: your name, that you are vegetarian, that you prefer concise answers. Think of it as a small notepad ChatGPT keeps about you. This notepad has a hard size cap, and this is what fills up.
  • Chat history reference is different: with it turned on, ChatGPT can pull relevant details from your past conversations. OpenAI states there is no storage limit on what it can reference this way.

So when ChatGPT says memory is full, it means the saved-memories notepad has hit its cap. OpenAI does not publish the exact size, but community measurements consistently put it at roughly 1,200 to 1,500 words in total, which works out to somewhere around 100 to 200 short entries. It is genuinely small — closer to a sticky note than a filing cabinet.

What it does not mean: nothing is deleted, nothing breaks, and your conversation history is untouched. ChatGPT keeps working and keeps using the memories it already has. It simply stops saving new ones until you make room.

One more thing worth knowing in 2026: OpenAI is rolling out a newer memory system (a continuously updated "memory summary") that manages itself more automatically, starting with Plus and Pro users in the US. If your Settings look different from screenshots in older guides, that is why. The advice below covers the saved-memories system, which is what produces the memory full message.

Step 1: See what is actually in there

Before deleting anything, look at what ChatGPT has been saving. It is usually an odd mix of the useful and the accidental.

  1. Click your profile icon and open Settings.
  2. Go to Personalization, then the Memory section.
  3. Click Manage memories to see the full list.

Expect to find entries you never consciously created. ChatGPT saves details on its own when it judges them useful for later, so a one-off role-play ("act as a pirate for this email") can end up living next to genuinely useful facts about your work. That accidental clutter is usually what fills the notepad.

Step 2: Free up space

You have three built-in moves, in increasing order of effort:

Delete the dead weight

In Manage memories, click the next to any entry and select Delete. You can also just tell ChatGPT in a chat: "Forget that I was planning a trip to Lisbon." Either way the memory stops being used in future chats. There is a Delete all option too, but it is rarely the right move — the notepad usually contains a handful of entries worth keeping.

Consolidate overlapping entries

ChatGPT can update, combine, or rewrite its own memories when asked. If you see five separate entries about your job, say something like: "Look at your saved memories about my work and combine them into one concise memory, then delete the originals." Several sentences become one, and you reclaim most of the space without losing anything.

Let it manage itself (Plus and Pro)

On Plus and Pro plans (web), OpenAI now offers automatic memory management: ChatGPT keeps the most relevant memories "top of mind" and moves less important ones to the background, specifically to help avoid the memory full state. You can search and sort saved memories, prioritize or deprioritize specific ones, and even restore earlier versions. If you are on a paid plan and tired of manual pruning, this is the low-effort answer.

Two quirks worth knowing: deleting a chat does not delete the memories saved from it — they live separately, so remove both if you want something fully gone. And turning off "Reference saved memories" in Settings also turns off "Reference chat history," which is a bigger change than most people intend.

What is worth keeping

Since the space is small, treat it like a small space. Memories earn their spot when they are stable and broadly useful:

  • Who you are and what you do, in one line.
  • Standing preferences: tone, format, units, language.
  • Long-running constraints: dietary needs, accessibility requirements, the tech stack you work in.

Almost everything else — project details, one-off contexts, things that were true for a month — is better off deleted. Those details date quickly, and a stale memory is worse than no memory: it quietly nudges answers in outdated directions.

The honest part: what clearing memory will not fix

Here is what most articles about this message skip. You can prune the notepad perfectly and two things remain true:

  • It will fill up again. ChatGPT keeps saving memories on its own. Pruning is maintenance, not a cure. (Automatic memory management helps, but the cap itself does not move.)
  • The notepad was never your conversations. Saved memories are distilled preferences — "prefers concise answers," "works in marketing." The actual substance of your work with ChatGPT — the plan you built across three long threads, the draft you refined for a week, the explanation that finally clicked — is not in there and never was. It lives in your chat history, in someone else's product, under their retention rules, findable only through their interface.

That is the real shape of the problem. The memory full message is a small annoyance about a small notepad. The bigger issue is that the valuable part — the conversations themselves — was never saved anywhere you control.

The real fix: keep your own copy of the conversations

I want to be precise about what I am claiming here, because this matters: ThreadRecall, the app I build, does not increase ChatGPT's memory cap. Nothing can. It solves the bigger problem underneath it.

ThreadRecall is a menu bar app for Apple Silicon Macs (macOS 13 or later, currently a free beta) that captures your AI conversations as they happen and keeps them on your machine:

  • Full conversations, not distilled notes. It captures conversations from the ChatGPT desktop app — and the Claude, Gemini, Codex, and Perplexity desktop apps — into a local SQLite database. The whole thread, not a one-line summary of it.
  • Local by design. Nothing leaves your Mac. No cap to hit, no notice asking you to delete your own history to make room.
  • Real search. Keyword search plus optional semantic search, so you can find "that thread where I worked out the pricing" without remembering the exact words.
  • Recall where you work. Inside Claude, trecall: commands pull past context back into the current conversation — including context that started life in ChatGPT.
  • Obsidian sync. If you keep notes, conversations can flow into your vault as structured Markdown.

Prune the notepad — it takes two minutes and it is worth doing. But the lasting fix for AI memory is not managing a 1,500-word allowance inside someone else's app. It is keeping the conversations themselves, on your own machine, where no message can ever tell you they are full.

Your conversations deserve better than a notepad.

ThreadRecall captures your ChatGPT, Claude, Gemini, Codex, and Perplexity conversations locally on your Mac and makes them searchable. Free beta, early access open.

Join early access

Frequently asked questions

Does the memory full message delete my chat history?

No. It only affects saved memories, the small notepad of facts and preferences. Your conversation history is stored separately and stays exactly where it was.

How many memories can ChatGPT store?

OpenAI does not publish an exact number. Community measurements put the cap at roughly 1,200 to 1,500 words total, which lands somewhere around 100 to 200 short entries depending on their length.

Does ChatGPT stop working when memory is full?

No. Conversations continue normally and existing memories are still used. ChatGPT just cannot save new memories until you free up space.

Can I increase ChatGPT's memory limit?

No setting raises the cap. Plus and Pro users can turn on automatic memory management, which prioritizes relevant memories to help avoid the full state, but the limit itself stays. No third-party tool raises it either — ThreadRecall included. Keeping full local copies of your conversations is a different (and, I would argue, better) kind of fix.

Does deleting a chat also delete the memories saved from it?

No. Saved memories live separately from chat history. To fully remove something, delete the saved memory in Manage memories and delete the chat where you shared it.