Claude to ChatGPT Transfer: How to Move Your Context Across

Search for how to move between these two and you will find the same article twenty times, all of it going the other way. ChatGPT to Claude is well covered, by Anthropic and by most of the tech press.

Claude to ChatGPT is not covered at all, and it is the direction more people are quietly taking.

Why the tooling only runs one way

Anthropic built an import page. You can paste a summary of what another assistant knows about you at claude.com/import-memory, or find it under Settings, Capabilities, Memory Import. There is no equivalent going the other direction, and there is no commercial reason for either company to build one.

So the honest starting point is that this is a manual move. That is not a failure on your part, and it is worth knowing before you go hunting for a button that does not exist.

Two things worth knowing about the Claude side before you start. Its memory import handles memory and preferences, not your chat history, Projects, or custom instructions. And long term memory on Claude requires a paid plan, so if you have been on free, there may be less stored than you assume.

What actually needs to move

The instinct is to move everything. Almost nobody needs to.

Your chat history is a record of how you got somewhere. What the next assistant needs is where you ended up.

Four things do the work:

The decisions you have settled, and the reasoning behind each one. Without the reasoning, the new assistant reopens questions you closed weeks ago.

The constraints. Formats, tone, technical limits, the things you have told Claude never to do. This is the part people leave out, and it is the reason a switch can feel wrong even when all the facts arrived.

The current state of anything in progress, including the latest version of the actual output.

What did not work, and why. This is the highest value and the least transferable thing you have. It is the only part of your history that cannot be reconstructed from the output.

Everything else is retrievable later. These four are not.

Getting it out of Claude

Ask, do not export. An export gives you a large file that ChatGPT will not read usefully.

Open the Claude conversation that matters and ask it to write a handoff covering those four things by name, plus the next step. Ask for it as plain text, not as a table or an artifact, because you are going to paste it into a different product.

Read it before you use it. Summaries drop constraints more reliably than they drop facts, and Claude will happily compress away the rule you spent an hour establishing. Add back what is missing. This takes two minutes and it is the difference between a switch that works and one that leaves you re-explaining everything anyway.

Do this per project, not for your whole account. One combined document covering four unrelated projects is worse than four short ones.

Getting it into ChatGPT

Two options, and they do different jobs.

For a one time move, paste the handoff as the first message of a new chat and work from there. Simple, and it dies when that conversation ends.

For something ongoing, put the durable parts into custom instructions instead. Preferences, tone, formats, standing rules. Keep the project specific parts in the chat. Mixing them is why people end up with custom instructions that are three paragraphs long and no longer describe how they actually work.

Either way, before asking for any work, get ChatGPT to state back what it thinks the goal and the constraints are. You will find what got lost immediately, and it is much cheaper to find it now than three replies into a task.

What will feel different anyway

Facts transfer. Behavior does not.

Tone, formatting habits, how much it hedges, how it structures an answer, all of that was partly Claude being Claude and partly months of you correcting it. None of that comes across in a handoff, and the same thing happens in every direction between any two models.

Expect to spend a week re-establishing it. Writing down the corrections you find yourself making, as they happen, shortens that considerably.

If you go both ways regularly

Plenty of people are not switching, they are using both, and doing this move every few days in both directions. At that point the per project handoff stops being a task and starts being a chore, and it is one of the things people name as a reason they gave up and went back to one tool.

The manual method does not scale past a couple of moves a week. If that is you, the question changes from how to move context to where your context should live so that moving it is not a project each time.

Where Unibase Memory fits

Unibase Memory is a Chrome extension that captures your conversations from Claude, ChatGPT, Gemini, Grok, Kimi and DeepSeek into one searchable archive in your browser, along with web pages you save.

Practically, it changes the direction of the work. Instead of producing a handoff in Claude and carrying it to ChatGPT, you search your archive for the conversation where the decisions were made, select the specific messages, and send them into the ChatGPT input box. You edit them there before submitting. The same works in reverse, and across the other assistants too.

Two things it does not do. It does not import into ChatGPT's own memory, and it does not decide what matters. You still choose what goes, which is deliberate, because the four things above are the ones you would not want software guessing at.

Your conversations stay on your machine by default, and everything exports as a file whenever you want.

Add to Chrome: www.unibase.com/invite/BLOG