ChatGPT conversation too long? Continue in a new chat
By Shaig Khaligli · · 9 min read
When ChatGPT says a conversation is too long, that chat has used up the space it can work with, and the fix is a new chat. To keep your context, ask the old chat for a handoff summary (decisions, facts, open questions, files) and paste it as the first message of the new one. If the old chat won't take another message, hover an earlier message, choose More actions (⋯) → Branch in new chat, and ask for the summary in the branch. For work that runs over weeks, put the chats in a ChatGPT Project so every new chat starts with the same instructions, files and saved answers.
Quick disclosure: I build TidyChats, a browser extension for ChatGPT, Claude, Gemini and Grok, and it has a button for this. The built-in ways come first, because most of the time they're enough.
Why ChatGPT says the conversation is too long
Every model can only hold so much text at once. That space is called the context window, and it covers your messages, ChatGPT's replies, and the files and pastes you dropped in along the way. When a chat outgrows what ChatGPT allows, it stops you from sending more. The message people report on OpenAI's community forum reads "You've reached the maximum length for this conversation, but you can keep talking by starting a new chat" (forum thread, August 2026). An older wording, "This conversation is too long, please start a new one", still turns up too. I couldn't find an OpenAI help article for the message itself, so treat both wordings as what users quote.
How much fits depends on your plan and the model you pick. I'm not putting numbers here: third-party sites disagree with each other, and OpenAI changes models often.
This is also separate from your usage limit. Waiting for a reset won't make the chat shorter. If ChatGPT told you to wait until a certain time, that's a plan limit, which I cover in ChatGPT usage limits and resets. If you also use Grok, its limits work differently, with one weekly pool; see Grok usage limits.
You don't have to wait for the error to move on. I'd start fresh whenever the topic has drifted or the chat has collected several big pastes, because both fill the window fast.
Move to a new chat with a handoff summary
This is the method I'd use first, and it works on every plan. While the old chat still answers, send it something like this:
I'm moving this work to a new chat. Write a handoff summary I can paste there:
1. The goal and the current status, in 2-3 sentences.
2. Decisions we made, with the reason for each.
3. Key facts, numbers, names and constraints the new chat must know.
4. The latest version of anything we produced (outline, code, copy), in full.
5. Open questions and the next step.
6. Files I uploaded that the new chat will need.
Keep it under 600 words, except item 4.
Then open a new chat, paste the summary as your first message, and add one line on what to do next ("Continue with step 5"). Read the summary before you send it. The model wrote it from a chat it was already struggling to hold, so check that the decisions and numbers are right.
A few habits make this work better:
- Re-upload the files. Uploads don't follow you into a new chat. The summary can name them; you still have to attach them again.
- Keep the latest version, not the history. If you went through four drafts, the new chat needs draft four, not the story of how you got there.
- Save the summary somewhere. A note, a doc, or a Project source (more on that below), so you have it if the new chat also grows long.
If the chat won't take another message: branch from an earlier point
Once ChatGPT refuses new messages, you can't ask that chat for a summary. On the web there's a way around it. OpenAI added branching in September 2025: hover over a message, click More actions (⋯), and select Branch in new chat to "start a separate conversation from that point" (ChatGPT release notes, available to all logged-in users on web).
Pick a message a bit before the end and branch from it. The branch only carries the conversation up to that point, so it's shorter than the chat that hit the wall, and you can ask it for the handoff summary above. You lose whatever came after the branch point, so paste the last few exchanges in yourself if they matter.
You can also copy the parts you need from the old chat by hand. Some users on the forum report that very long chats sometimes won't open or export at all (forum thread, May 2026), and OpenAI's support account replied there asking people to file a ticket. One more reason to move on before a chat gets that far.
Use a Project for work that keeps going
For work you'll come back to (a thesis, a client, a codebase), ChatGPT Projects keep related chats, files and instructions together. OpenAI's Projects help article describes what helps here:
- Move the long chat in. Drag it onto the project, or open the chat's menu and select Move to project. A moved chat picks up the project's instructions and file context. Chats made with a GPT can't be moved.
- Write project instructions. Open the project's ••• menu → Project settings. They apply only inside that project and override your global custom instructions.
- Save answers as sources. In a project chat, open the menu on a response and choose Add to project sources. Your handoff summary is a good candidate: every future chat in the project can use it.
- Let project chats reference each other. Project chats can draw on other conversations in the same project, as long as memory is on (for most plans that means Reference saved memories and Reference chat history). For Plus and Pro, OpenAI says ChatGPT prioritizes the project's chats and files when you ask a question there.
Files per project depend on your plan: 5 on Free, 25 on Go and Plus, 40 on Pro, Business, Enterprise and Edu, as of November 2026.
My take: a Project doesn't make one chat longer, but it makes starting the next chat cheap, which is the next best thing.
What memory does (and doesn't) carry over
ChatGPT's memory lives in Settings → Personalization → Memory. With Reference chat history on, ChatGPT can use relevant information from past chats to personalize new ones. That sounds like the answer, but OpenAI's memory article is clear about its limits: memory "does not retain every detail from every conversation", ChatGPT decides what's relevant, and it doesn't search your history for every response.
So memory is good at "I prefer metric units" or "I'm writing a fantasy novel set in Lisbon". It's not a reliable way to bring chapter outlines, code or a list of decisions into a new chat. For working context, use a summary or a Project.
Which way carries what
| Method | What the new chat gets | Good for |
|---|---|---|
| Handoff summary | What you paste: decisions, facts, latest drafts | Any plan, any one-off move |
| Branch in new chat | The conversation up to the message you pick | A chat that already hit the limit (web) |
| Project | Instructions, files, saved sources, other project chats | Ongoing work you return to |
| Memory | Preferences and details ChatGPT picks as relevant | Personal background, not working files |
Make the next chat last longer
- One topic per chat. Start a new chat when the subject changes instead of keeping one thread alive for a week.
- Attach big texts instead of pasting them. ChatGPT now turns pastes over 10,000 characters into an attachment on all plans, which OpenAI says helps keep a big paste from using up your context window. You can switch it back with Show in text field.
- Check the size first. The token counter estimates how big a document is before you paste it.
- Plan to find old chats. Splitting work across chats makes them harder to find later; here's how to search your ChatGPT history.
- Don't babysit long answers. If you switch tabs while ChatGPT thinks, here's how to get notified when ChatGPT finishes.
Continuing a chat with TidyChats
TidyChats does two things for this. First, it shows how full the chat is before ChatGPT stops you. A thin meter under the chat box reads "N% of chat memory used · N tokens". It's gray under 50%, amber from 50% and red from 80%. Click it to see this chat's tokens and the model memory it's measured against, plus a note that long chats make the AI forget the beginning.
Second, the same panel has Continue in a new chat. It's also on Home in the TidyChats window as Continue in another AI, and in the command palette. It summarizes the chat you're in and opens a new chat, on ChatGPT or on Claude, Gemini or Grok, with the summary in the chat box. On ChatGPT it also attaches the images and documents you uploaded in the chat (up to 10 files, 20 MB) and lists their names; anything it couldn't carry is named too. Nothing is sent for you: you read the summary, wait for the upload, and press send.
If you'd rather pull an old chat into a new message, type @@ in the chat box and pick a past conversation. TidyChats inserts an extract of up to about 1,200 words. The @@ context mentions tutorial walks through it.
The honest limits:
- The meter is an estimate. It doesn't read ChatGPT's real limit. When ChatGPT says the chat is too long, ChatGPT is right.
- The free summary is built in your browser. In my view it won't be as tailored as the handoff prompt above, because it doesn't know what you plan to do next. Read it and add a line before sending.
- AI summaries are Premium. For signed-in Premium users, the chat text (up to 60,000 characters) goes to the TidyChats server to be summarized. Everyone else gets the local summary, which isn't sent to our server.
- @@ on Free gives 3 mentions a day per AI, 1 per message, from chats on the same AI. Premium removes the cap and allows chats from another AI.
- Gemini and Grok chats carry the summary only, no files.
Premium is $6.99 a month for all four assistants. More on the TidyChats page for ChatGPT and the pricing page.
Questions people ask
Why does ChatGPT say my conversation is too long?
The chat has reached the most text ChatGPT will handle in one conversation, counting your messages, its replies and anything you pasted or uploaded. The fix is a new chat with a summary of the old one.
Can I make a ChatGPT chat longer?
No setting does that. Plans and models differ in how much they can hold, but every chat has a limit. Keep topics in separate chats, attach long texts instead of pasting them, and move on with a summary when the chat gets heavy.
Does a new chat remember the old one?
Only partly. With Reference chat history on, ChatGPT may use relevant details from past chats, but OpenAI says memory doesn't keep every detail. Inside a Project, chats can draw on the project's other chats, files and sources. For anything specific, paste a summary.
How do I continue a chat that has already hit the limit?
On the web, hover an earlier message, choose More actions (⋯) → Branch in new chat, and ask the branch for a handoff summary. Then start a fresh chat with that summary, or copy the parts you need from the old chat by hand.
Do ChatGPT Projects fix the length limit?
Not for a single chat. What a Project fixes is the restart: new chats there start with the project's instructions, files and saved sources.
Try TidyChats free
Folders, search and prompts for ChatGPT, Claude, Gemini and Grok.