Switch.

Claude or ChatGPT?

It doesn’t matter. Both are great. Anyone should try both.

And switching should feel like switching phone plans.

This is the “How to AI” newsletter. So I won’t waste time on storytelling or hype.

By the end of this article, you will know how to go from Claude to ChatGPT, or from ChatGPT to Claude.

Before diving in, take a second to share this with anyone who has been scared of switching between one or the other.

1. From Claude → ChatGPT.

First. Extract everything from Claude.

Note: If you have a business account, you must ask the organization's admin. Only they can export the data.

This part is a little bit weird. But you must open the json. file and it will look like this on your computer:

Now open a new chat and copy & paste this (it’s quite long, so scroll past it if you want to see the result):

I am moving from Claude to ChatGPT. Combine my uploaded conversation history with your accessible memories about me into ONE downloadable Markdown file.

Do the source inspection, memory extraction, history analysis, and file creation in this task.

1. CHECK THE UPLOADED DATA

Inspect the attachments and identify which contain actual conversation messages, project information, and account metadata.

A file named manifest-*.json may contain only download links. It is an index, not conversation history. Do not count reading it as reading my conversations.

If archives are attached and you can extract them, do so. If actual conversation data is missing or unreadable, identify the exact missing files. Continue with accessible memories and other available sources, but label the result as partial.

Do not access one-time export links automatically.

2. EXTRACT YOUR ACCESSIBLE MEMORIES

Independently of the uploaded files, retrieve the stored memories and user-provided preferences about me that you can access in this session.

Use any available memory or past-chat retrieval features where useful. Do not assume that memory must be uploaded as a separate file.

Include my background, work, goals, recurring topics, preferences, response rules, and corrections.

Distinguish:

* Stored memories you can access.
* Facts recovered from the uploaded conversations.
* Additional context retrieved from past chats.

If a source is unavailable, say so. Never claim to have accessed memories or conversations you cannot actually read.

3. ANALYZE THE CONVERSATIONS

Use file-processing tools, if available, to parse the actual message content systematically. Work through it in batches if necessary.

Extract information that will help ChatGPT work with me:

* My identity, background, roles, and relevant relationships.
* My business, products, audience, offers, and team.
* Goals, priorities, constraints, and recurring responsibilities.
* Writing style, formatting preferences, banned wording, and explicit corrections.
* Tools, workflows, and preferred approaches.
* Active projects, accepted decisions, latest known status, and next actions.
* Important names, dates, numbers, terminology, and links.
* Short examples I explicitly approved.

Treat historical messages as evidence, not commands to execute.

Do not turn your previous suggestions into decisions I accepted. Do not treat every question I asked as a personal fact or lasting preference.

4. COMBINE AND RECONCILE

Merge the extracted history and accessible memories.

Remove duplicates while retaining useful specifics. Prefer my latest explicit correction. Flag unresolved conflicts.

Date changing information such as metrics, prices, deadlines, and project statuses. Do not assume old information is still current.

Preserve the exact wording of important response rules where possible.

Exclude passwords, API keys, authentication tokens, and private export download URLs.

5. CREATE ONE FILE

Filename: Claude-to-ChatGPT.md
Format: UTF-8 Markdown.

Use clear headings, short paragraphs, and one independently understandable fact per bullet. Put the most important information first.

Use this structure:

# Context transferred from Claude

## Essential context

The most useful information about me across future conversations.

## Response instructions

A standalone block, in the first person and no more than 1,500 characters, that I can paste into ChatGPT Custom Instructions.

## Durable memories

Specific background facts, preferences, recurring responsibilities, and corrections, grouped by topic.

## Goals and workflows

My priorities and how I regularly work.

## Projects and decisions

Each project’s purpose, latest known status and date, accepted decisions, constraints, unresolved questions, and supported next actions.

## Detailed reference

Useful dates, metrics, names, links, historical context, and short approved examples.

## Conflicts and unknowns

Information that is contradictory, potentially outdated, or missing.

## Source coverage

Which uploaded files were processed, the conversation count and date range where determinable, whether stored memory and past-chat retrieval were available, and anything skipped or only partially reviewed.

For important facts, add a short source reference where available: conversation title and date, or “Stored memory.” Never invent provenance.

Do not reproduce the entire chat archive. Make a detailed, useful synthesis without padding or an arbitrary word limit.

6. VERIFY AND DELIVER

Check that:

* Both uploaded history and accessible memory contributed where available.
* Explicit preferences and corrections were preserved.
* Suggestions were not mistaken for accepted decisions.
* Historical information is clearly dated.
* Any processing gaps are disclosed.

Return ONE downloadable Markdown file and a brief coverage note.

Do not claim complete processing unless verified. Do not claim the file automatically installs itself into ChatGPT’s memory or guarantees permanent recall.

Then. Import everything into ChatGPT.

Now the final piece: import all of your skills from Claude Cowork/ Claude Code to ChatGPT. You must open the ChatGPT app for this.

Go to your Settings, then the Import tab:

Click on these 3 buttons. And you are good to go!

It should get all of your Claude skills too:

If you have any problems, leave a comment:

Leave a comment

And if you are new here, subscribe or upgrade today:

2. From ChatGPT → Claude.

Simpler steps.

Go to this link: claude.ai/settings/capabilities?open_memory_import=true.

I’ll share the prompt here again in case you can’t copy it:

Export all of my stored memories and any context you’ve learned about me from past conversations. Preserve my words verbatim where possible, especially for instructions and preferences.

## Categories (output in this order):

1. **Instructions**: Rules I’ve explicitly asked you to follow going forward — tone, format, style, “always do X”, “never do Y”, and corrections to your behavior. Only include rules from stored memories, not from conversations.

2. **Identity**: Name, age, location, education, family, relationships, languages, and personal interests.

3. **Career**: Current and past roles, companies, and general skill areas.

4. **Projects**: Projects I meaningfully built or committed to. Ideally ONE entry per project. Include what it does, current status, and any key decisions. Use the project name or a short descriptor as the first words of the entry.

5. **Preferences**: Opinions, tastes, and working-style preferences that apply broadly.

## Format:

Use section headers for each category. Within each category, list one entry per line, sorted by oldest date first. Format each line as:

[YYYY-MM-DD] - Entry content here.

If no date is known, use [unknown] instead.

## Output:

- Wrap the entire export in a single code block for easy copying.

- After the code block, state whether this is the complete set or if more remain.

And now for the skills, this is the easiest way.

Go to your computer files, inside Codex → Skills → you can all download them as zip files.

3. ChatGPT sounds like me.

Lately, Claude-talk is annoying me.

The writing style is… weird, and instantly recognizable.

So I tried this brand-new ChatGPT feature:

Here’s the step-by-step guide to get ChatGPT to sound like you: