ChatGPT
ChatGPT handover guide

How to Continue a ChatGPT Conversation in a New Chat

To continue ChatGPT in a new chat, copy a short handover of the goal, active instructions, decisions and latest work. Paste it as the first message, supply any essential files, and state the next task. If the old conversation has reached its limit, use the manual template below with the text you can still access.

Already have an export? Recover one conversation →

If ChatGPT shows “Chat memory full”

“Context window full” and “Chat memory full” describe how much information ChatGPT can use in this conversation. If ChatGPT offers to continue in a new chat with a summary, review that summary before relying on it: OpenAI says it may omit earlier instructions or exact wording. This is native continuation, not a transcript import.

A new chat does not erase saved chat history or saved Memory, but it does not automatically contain the old transcript, its exact working context, or files attached only to that conversation. Reattach needed files and paste important passages. A new chat also does not reset separate upload or account usage limits.

Save important text and files before switching. If the summary is incomplete or the option is unavailable, use the reviewed manual handover below. A chat-length notice alone is not a reason to delete saved memories or chat history.

Checked 14 November 2026 against OpenAI’s current guidance on full chats, GPTs, memory, and file uploads. Why ChatGPT says a chat is full · OpenAI Memory documentation · File upload limits

Use a GPT in this chat, or start a separate GPT chat?

On ChatGPT web, type @ in an existing conversation and choose an available GPT to send it your next message; the conversation keeps its current context. To begin a separate conversation with a GPT, open that GPT from the GPTs menu. Neither action imports an old transcript into a different chat. OpenAI says @ mentions are not available in its iOS or Android apps; use the GPT menu there. See OpenAI’s GPT guide.

How to continue a ChatGPT conversation in a new chat

  1. 1

    Open the original conversation

    Find the chat you want to continue and identify the work that must carry over.

  2. 2

    Create a structured handover

    Write one manually, or use thredly in the web app or optional Chrome extension to create one from the conversation.

  3. 3

    Review the handover

    Check that key constraints, decisions, current outputs, and anything ChatGPT must not repeat are included.

  4. 4

    Start a fresh ChatGPT chat

    Open a new conversation rather than adding a large transcript to the old one.

  5. 5

    Paste the handover first

    Make the handover the first message so the new chat has the working context before it starts.

  6. 6

    State the exact task to continue

    End with a direct next action, such as “continue drafting the implementation plan from section three.”

Manual ChatGPT handover template

A manual handover can be enough when you need to move quickly. Copy this template, fill in only what matters, and paste it into the new chat.

Copy this template, then replace each prompt with the current details from your conversation.

Objective
What I am trying to achieve:

Important background
The context the next chat needs:

Instructions and constraints
Rules, preferences, requirements, and things to avoid:

Decisions already made
What was chosen, and why:

Work completed
Useful drafts, code, research, or outputs so far:

Current version or output
The latest state to work from:

Open issues
Questions, risks, or missing information:

Next action
Continue by doing this:
Build a free manual handover Fill in the essentials, then copy or download

Write your next chat’s working brief

This builder formats text in your browser. It does not upload or save your entries. Copy or download before leaving. It does not check your sources or use AI.

Objective and next action are required. Add other details only when they matter; empty optional sections are omitted. Each field accepts up to 10,000 characters.

Your handover details

What are you trying to achieve?

What rules or boundaries still apply?

What was decided, and why? Mark proposals as unconfirmed.

What is complete? Name the latest draft, file or version.

What is missing, uncertain or unresolved?

What exactly should the next chat do?

Have a long source conversation? thredly can optionally draft an automated summary for you to review. Try an automated summary. Free account required; this manual builder stays free without an account.

What makes a handover useful?

Treat it as a working brief, not a transcript. Keep the details that change what the next chat should do: the goal, the current version, decisions that should not be reopened, and the task that remains. Leave out discussion that no longer affects the work.

Check it before you paste

  • Could someone understand the objective without opening the old chat?
  • Does it name the latest output or material to work from?
  • Does it finish with one specific next action?

In the new chat, paste the handover first. Then add a short instruction such as: “Use this handover as the working context. Continue by reviewing the open issues and drafting the next section.” This makes the immediate task clear without asking the new chat to reconstruct the old discussion.

If the work depends on a document, draft, code file, or research result, name the current version and explain what the next chat should do with it. If you move to a different AI, repeat any important format, tone, or scope requirements rather than assuming they will carry over from the old conversation.

Accuracy matters more than length. Remove stale instructions, confirm the current decision when the conversation changed direction, and keep the final request in plain language. That gives the new chat a dependable starting point and gives you a clear record of what was handed over.

Ask ChatGPT to create the continuation summary

If the original chat still opens, ChatGPT can produce a useful first draft of the summary. Copy this prompt into the old conversation, then review the result before using it in a new chat.

Copy this prompt into the original conversation, then check the summary before moving it to a new chat.

Create a continuation summary that I can paste into a new ChatGPT chat.

Base it only on this conversation. Use these headings:

- Objective
- Essential background
- Active instructions and constraints
- Decisions made and why
- Work completed and current output
- Unresolved questions and actions
- Superseded information to ignore
- Exact next action

Be explicit about anything uncertain or missing. Do not invent details. Keep the summary concise enough to use as the first message in a new chat.

Check the handover against your sources

Try the built-in continuation options described above first. If you still need a separate brief, you can write and check it yourself. Keep the original text beside you: a smooth summary can still contain the wrong name, an old date or a decision nobody approved.

  1. Save the readable messages and latest work you need. Label useful excerpts S1, S2 and so on. Keep sensitive details out unless the next task needs them.
  2. For each detail that changes the next task, fill in the worksheet below. Copy names, dates, times, units and amounts precisely. Keep an unspecified year or time zone marked unknown.
  3. Find the explicit correction or approval before treating a newer statement as settled. Record which earlier choice it replaces. Keep suggestions separate from decisions.
  4. Check every current claim against its source. If the source is missing, say so. Ask the next chat to resolve essential gaps before doing more work.
Task to continue:
Latest output or version:
Sources supplied:
Sources missing:

Repeat for each important fact, instruction or decision:
- Current detail:
- Status: confirmed / proposed / rejected / superseded / unknown
- Source ID and exact supporting words:
- Earlier detail replaced, if any:
- Correction or approval source:

Open questions:
Exact next action:

Read this brief and the supplied excerpts. Restate the next task and flag conflicting or missing information before continuing. Keep proposals separate from confirmed decisions. Do not invent details or reopen a rejected choice without asking.

This worksheet checks the material you have. It cannot recover inaccessible messages or prove that missing context has been preserved.

Old chat won’t respond? Recover one conversation from an export.

If you can still read the chat, copy the important messages or use the manual handover above now. If copying is awkward and you already have a downloaded export, thredly’s local importer can help you select one conversation and one branch before preparing a handover.

A downloaded ZIP can contain sensitive account data beyond the chat you need. Keep it private; do not send it to us.

Need an export first? Allow up to 7 days.

In an eligible ChatGPT account, open Settings → Data controls → Export data and confirm. OpenAI says delivery can take up to 7 days and the download link expires after 24 hours. Self-service export is unavailable in Business, Enterprise and Healthcare workspaces; ask the workspace owner about managed data access. An export cannot restore deleted chats.

OpenAI export instructions · checked 29 October 2026.

Check file limits: large exports can fail before you select a chat.

Large account exports can fail before you get to choose a chat: selection happens after the files are read and parsed. Limits include 25 MiB per archive, 8 MiB per text/JSON file, 16 MiB of combined text/JSON leaves, 50 MiB of selected source files and 20 independently readable parts. Archive expansion has a separate 100 MiB limit, including skipped TGZ media.

If the archive is too large, you can extract an existing supported conversations JSON file on your device and check its size first. A JSON file containing many chats can still exceed the limits. Do not rename or chop archive bytes into smaller files. If no supported file fits, use readable messages, saved source files and the manual template instead.

  1. Choose the source. Sign in to the thredly web app, choose Import file, and select the downloaded file or independently readable export parts. TXT/Markdown, JSON, JSONL/NDJSON and supported text exports inside ZIP/TAR/TGZ are available. HTML, media and nested archives are skipped; binary split archives are unsupported.
  2. Choose one conversation and branch. In Import text, use Choose a conversation or file. For a recognized ChatGPT structure, check Conversation branch: Current branch follows its recorded current node; alternative replies stay separate. The adapter recognizes supported ChatGPT export structures, not every export. If it cannot recognize yours, review explicit JSON fields or use a readable text copy; never guess which metadata is the conversation.
  3. Preview before applying. Choose Preview selected text. Check the first and last messages, speaker order, current decisions and omission warnings. Open essential attachments separately: importing text does not carry the files into a handover.
  4. Replace or Append deliberately. With an existing draft, choose Replace existing text or Append after existing text, then check the combined preview. Confirm with Replace with this text or Append this text. With an empty draft, Use this text fills the composer. Importing does not submit a summary.
  5. Create, then review. Once the composer contains the intended text and meets your account’s limits, choose Create summary explicitly. Review the resulting handover against the source, restore missing constraints, and give the fresh chat the exact next task. File reading and selection stay in the local importer; applied text follows the web app’s existing draft and explicit summary-submission behavior.

Worked selection: one plan, not every chat

This hand-authored synthetic export contains Booking-page launch plan and the unrelated Weekend walking route. The plan has a current branch and an alternative reply proposing payments. Choose Booking-page launch plan (ChatGPT) and Current branch.

Download the synthetic export (JSON)

Expected selected text below: verified with the existing importer against an independently specified text fixture. This is extracted text, not a thredly summary or a customer result.

User:
Continue the booking-page pilot. Scope: enquiry form only; no payments or reminder emails. The project owner is Priya.

Assistant:
Current plan: keep the enquiry form. Priya will check field labels; Morgan must approve retention wording. Launch date is not agreed.

User:
Next action: prepare the label checklist. Check the source file separately; the attached mockup is not part of this text.

The walking route and superseded payments reply are absent. Selecting Alternative branch 1 instead includes that alternative reply; it does not merge it with the current plan. A non-text mockup marker is omitted with a warning. Check the real source file separately.

This shortened example is below the 5,000-character summary minimum. Short text can be imported, but cannot be submitted as a summary yet. Use the manual template for short notes; do not pad them. A free account’s summary accepts up to 500,000 characters; other account limits still apply.

Ways to continue or preserve a ChatGPT conversation

ChatGPT now has several built-in options, but they solve different problems. Branching explores another direction from a chosen message. Projects keep related chats, files, and instructions together. Memory can use relevant context from past chats when enabled, but it does not retain every detail. A data export provides an account archive rather than a ready-to-paste continuation summary.

Reviewed 2 September 2026 against current OpenAI documentation for branching, Projects, Memory, and data export.

Comparison of seven ways to continue or preserve a ChatGPT conversation
MethodBest forWhat carries forwardMain limitation
Manual summaryA short or sensitive handover you want to controlOnly the details you select and verifyTakes time to write and check
ChatGPT-generated summaryCreating a quick first draft from a chat that still opensThe details ChatGPT identifies from the current conversationMay omit constraints, decisions, actions, or stale directions unless reviewed
BranchingExploring a different direction from a particular messageThe conversation up to the branch pointStarts another path with the earlier context; it is not a condensed handover
ProjectsKeeping related chats, files, and project instructions togetherProject chats, uploaded sources, and project instructions, subject to settingsOrganizes ongoing work rather than summarizing one long conversation
Memory / reference historyPersonalization and relevant details across chatsInformation ChatGPT decides is useful when Memory is enabledDoes not retain every detail and should not replace a task-specific brief
Data exportKeeping an archive of eligible account data and chat historyA downloadable account archiveMay take up to seven days and is not formatted as a continuation prompt
thredly structured handoverTurning one long conversation into an inspectable brief for a fresh chatGoals, constraints, decisions, progress, outputs, open issues, and next stepsYou still need to review and paste the handover; it is not automatic cross-chat sync

If you prefer not to assemble the brief yourself, thredly can create a structured handover for you to inspect, edit, and paste after you have tried the free manual or ChatGPT-generated approaches.

Example of checking changed travel plans

Invented teaching example. The excerpts, people and trip are fictional. The flawed recap is written to illustrate mistakes; it is not a measured AI result.

Original excerpts

  1. S1 · User

    Plan a Lisbon weekend for Jo and me, 16–18 October 2026. Aim for a hotel at €180 per night or less. Nothing is booked.

  2. S2 · Assistant

    One option is to arrive at 18:00 on 16 October and choose Hotel Azul.

  3. S3 · User

    Correction: my companion is Mo, not Jo. Change the trip to 23–25 October 2026. Keep the €180-per-night hotel budget. Hotel Azul is only an option; do not book it.

  4. S4 · User

    The flight times are still undecided. Compare two hotel options next. Leave transport planning until we choose our flights.

A recap worth rejecting

Jo and I are arriving on 16 October at 18:00. Hotel Azul is booked within our €180 total budget. Plan the airport transfer.

Checked handover

Task: compare two hotel options for the Lisbon trip.
People: user and Mo [confirmed: S3, "my companion is Mo, not Jo"]. Jo is superseded [S1 → S3].
Dates: 23–25 October 2026 [confirmed: S3]. The 16–18 October dates are superseded [S1 → S3].
Hotel budget: €180 per night or less [confirmed: S1 and retained by S3]. This is not a total-trip budget.
Hotel Azul: an option only; booking is not authorized [proposed: S2; clarified by S3]. No booking confirmation is supplied.
Flight times: undecided [S4]. The suggested 18:00 arrival is not confirmed [S2]; no time zone is supplied.
Latest output: the hotel suggestion in S2; no comparison has been completed in these excerpts.
Next action: compare two hotel options. Leave transport planning until flights are chosen [S4].

The useful corrections are small and exact: Mo, the later dates, a nightly budget and an unbooked option. Check current prices and schedules when you make real travel decisions.

Optional help drafting the brief

You can use the worksheet on its own. If you want help drafting a longer brief, see Create my ChatGPT handover above. Check the result against your sources before using it.

Stay in the existing chat or restart?

Stay in the current chat when

  • the conversation is still responsive and focused
  • recent messages contain details you are actively using
  • you are finishing a small step and a restart would interrupt it

Start a fresh chat when

  • the conversation is slow, inconsistent, or repeatedly losing context
  • old instructions conflict with the current direction
  • you need a clean brief for a new phase, collaborator, or AI

How thredly handles the handover

Turn the old ChatGPT thread into a handover you can inspect

The manual method above remains available. thredly automates the first draft, then leaves you in control of what the new chat receives.

  1. 1. Add the conversation

    Paste your ChatGPT conversation into the web app, or use thredly inside the supported chat.

  2. 2. Review the handover

    Check the objective, instructions, decisions, current work and next action.

  3. 3. Continue in a fresh chat

    Paste the reviewed handover into a fresh ChatGPT chat, then state the exact task to continue.

Why not paste the entire conversation?

Keep the full transcript as a record, but select the current decisions and instructions for the next task. Our AI chat portability case study shows why transferring text and reconstructing the current state are different. Check the handover against your source.

What thredly preserves

A thredly handover is designed to preserve goals, instructions, constraints, decisions, rejected approaches, current progress, important outputs, open questions, and next steps.

ChatGPT Memory versus a conversation handover

ChatGPT Memory can retain selected preferences or facts. It should not be treated as a complete replacement for transferring the working state of one specific long conversation. A handover is designed for continuing a particular piece of work: its goal, constraints, decisions, outputs, and next action. See the ChatGPT Memory guide, compare ChatGPT Memory with a thread handover, or review how Custom Instructions differ from continuation context.

What if the new chat still behaves inconsistently?

Treat the handover as a working brief you can correct. If the new conversation misunderstands it, stop and repair the context before asking for more work.

  1. 1

    Restate the objective and the exact next action in one short message.

  2. 2

    Add any missing constraint or decision that changes the answer.

  3. 3

    Remove stale instructions and name the final direction explicitly.

  4. 4

    Paste or attach the latest draft, file, or output if the task depends on it.

  5. 5

    If the brief contains several unrelated tasks, continue one task at a time.

Why long ChatGPT conversations break down

The longer a ChatGPT conversation runs, the harder it becomes for the model to keep track of everything you've discussed. What starts as a focused, productive thread gradually drifts.

  • Instructions you gave early on get buried under newer messages
  • Key decisions fade and ChatGPT stops referencing them
  • The conversation drifts away from the original goal
  • You start repeating yourself to remind ChatGPT what you already agreed on
  • Responses become more generic as the thread gets longer

At some point, the conversation stops being useful, not because you ran out of things to discuss, but because the context has degraded.

What to include in the summary

A good summary is not a transcript. It captures the parts of the conversation that the next chat actually needs to work with.

Instructions and constraints

Any rules, preferences, or boundaries you set during the conversation that should still apply.

Decisions made

Choices you and ChatGPT agreed on, architecture decisions, selected approaches, rejected alternatives.

Current progress

What has been completed so far, including any outputs, drafts, or partial work.

Next steps

What you were about to work on when the conversation started losing focus.

Why an in-chat summary still needs review

Asking ChatGPT to summarize the current thread can be useful. In a very long conversation, however, the first draft may emphasize recent messages or omit an early constraint that still matters.

  • Compare the summary with the current objective before moving
  • Restore any early instructions or constraints that are still active
  • Remove abandoned ideas that no longer affect the work
  • Name the latest output and finish with one exact next action

A reviewed handover gives you a chance to correct omissions before the new conversation treats the summary as its working brief.

Frequently asked questions

Create the handover before you start the new chat

Keep the important working context clear, review it once, and give the fresh chat a precise next action. Use the web app or the optional Chrome extension when you want thredly to create the structured handover.