How to switch from ChatGPT to Claude or Gemini: A Stress-Free 12-Step Guide

Why It’s Time to Move from ChatGPT to Claude or Gemini

The recent partnership between OpenAI and the Department of War has sparked a massive shift in how we view AI safety. While OpenAI has traded “hard” technical blocks for flexible military contracts, many users are rightfully concerned about the risks of AI hallucinations in high-stakes environments. If you believe that AI guardrails should be hard-coded rather than just written in a contract, you aren’t alone. This guide will walk you through how to migrate your workflows from ChatGPT to Claude or Gemini – platforms that prioritise rigorous ethical frameworks and technical neutrality.

Switching AI assistants can feel like moving house: you don’t want to lose anything important, and you definitely don’t want to unpack chaos. The good news is you can migrate smoothly if you treat it like a structured project.

This guide shows you how to move from ChatGPT to either Claude or Gemini in a way that preserves the value of your work: your best conversations, your reusable prompts, your tone and preferences, and the “brains” of your Custom GPTs – rebuilt as Claude Projects or Gemini Gems. Along the way, you’ll learn how to export official ChatGPT data, handle conversation JSON responsibly, and create portable “migration briefs” that any model can use.

Why People Switch: Cost, Capability, Privacy, and Workflow Fit

People usually migrate for one (or more) of these reasons:

• Workflow fit: You may prefer Claude’s project-based organisation or Gemini’s tight integration with Google tools.
• Different strengths: Some users like Claude for long-form reasoning and project organisation, while others prefer Gemini for Docs/Workspace workflows.
• Cost and plans: Pricing and plan bundles change, so many users diversify instead of betting everything on one platform.
• Data control: You may want a cleaner separation between work and personal chats, or different retention settings.

You don’t have to be “loyal” to a single assistant. A smart end goal is a setup where your knowledge is portable and your assistants are easy to recreate.

Before You Move: Define What “Migration” Means for You

Migration can mean different things:

1. Conversation archive (keep and search your history)
2. Working memory (your preferences, tone, recurring context)
3. Custom assistants (Custom GPTs → Gems / Projects)
4. Knowledge files (PDFs, docs, notes, playbooks)
5. Repeatable workflows (prompt templates, checklists, pipelines)

If you try to “move everything,” you’ll overwhelm yourself. The win is moving the 80/20: the conversations and assets that keep paying you back.

Quick Reality Check: What Transfers Cleanly vs. What Won’t

Here’s the honest truth:

What transfers well

• Your preferences and instruction style
• Your best prompts and templates
• Your knowledge files (documents you upload)
• Your conversation content as searchable archives (HTML/JSON/Markdown)
• Your Custom GPT “design” as a written specification

What doesn’t transfer perfectly

• A one-click “import my entire ChatGPT chat history into Claude/Gemini as native chats” (usually not available)
• Tool connections that are platform-specific
• Exact model behavior (different assistants respond differently)

So the goal is not “perfect recreation.” The goal is portable context plus rebuildable assistants.

Step 1 – Export Your ChatGPT Data the Official Way

ChatGPT provides an official export flow inside settings. The current standard method is:

• Open Settings
• Go to Data Controls
• Choose Export Data
• Confirm export and download the ZIP from your email link

That export is your “source of truth” backup. Even if you later decide not to migrate, you’ll be glad you did this.

What’s Inside the Export ZIP

Most exports include:

• A browser-readable file (commonly chat.html)
• A structured JSON export (often conversations.json) containing messages and metadata

Keep that ZIP somewhere safe (encrypted storage is ideal).

Step 2 – Triage and Organise: Decide What’s Worth Moving

Don’t migrate everything. Migrate what matters.

Create a simple “Migration Inventory” list with:

• Conversation title
• Topic (sales, marketing, coding, hiring, personal, etc.)
• Value score (1–5)
• Destination: Claude Project / Gemini Gem / Archive only
• Notes (contains sensitive data? contains reusable templates?)

Your goal: identify the top 10–30 threads that contain reusable knowledge (prompts, processes, checklists, writing frameworks).

Step 3 – Turn Long Chats into Portable Summaries

Long chats are messy. Models work best with clean briefs.

The trick is to turn any important thread into a Migration Brief that includes:

• What you were trying to do
• The key decisions
• The final outputs
• Your preferences and constraints
• “What to continue next”

A Copy-Paste Prompt That Produces Great Migration Briefs

Use this inside ChatGPT on the thread you want to migrate:

Prompt (copy/paste):

• “Summarise this conversation into a Migration Brief I can paste into another AI assistant. Include: (1) goal, (2) context, (3) constraints, (4) important decisions, (5) final outputs, (6) reusable templates/prompt snippets, (7) next steps, (8) my preferences for tone and formatting. Keep it under 500–900 words. Use headings.”

Then save the brief in a folder like:

• /Migration/Briefs/marketing-brief-2026-01.md
• /Migration/Briefs/product-strategy-brief.md

This single habit makes switching assistants dramatically easier.

Step 4 – Export / Recreate “Custom GPT” Behavior as a Spec

You usually can’t “convert” a Custom GPT automatically into another platform’s assistant. But you can recreate it reliably if you write an Assistant Spec.

Your spec should include:

• Name
• Purpose
• Personality / tone
• Rules (do/don’t)
• Inputs it expects
• Output format
• Examples (very important)
• Knowledge sources (files, links, reference docs)
• Safety boundaries (what it should refuse or avoid)

The “Assistant Spec” Template (Works Everywhere)

A simple structure:

1. Role: “You are a …”
2. Mission: “Your job is to …”
3. Audience: “You’re helping …”
4. Rules: bullets, prioritised
5. Process: step-by-step thinking style (without revealing hidden reasoning)
6. Output format: headings, tables, JSON, etc.
7. Examples: 2–5 real examples of input → output
8. Knowledge: what files or references matter
9. Limits: what it should not do

If you do this once, you can rebuild that assistant on any platform in minutes.

Step 5 – Move to Claude the Fast Way: Import Memory

Claude now offers a direct memory import flow designed for switching providers. The official “import memory” entry point is available at Claude’s import page, and the help documentation describes starting the import from Settings → Capabilities → Memory → Start import (or via a home screen card). 

What to Paste Into Claude’s Import Flow

Paste a clean, structured memory package, such as:

• Preferred writing tone (short, direct, friendly, formal)
• Formatting rules (bullets, tables, headings)
• Your standard workflow (e.g., “always ask for constraints, then propose 3 options”)
• Your business context (industry, audience, product positioning)
• “Always/never” instructions

This is where your Migration Briefs shine: they’re already formatted for exactly this.

Step 6 – Rebuild Custom GPTs as Claude Projects

Claude Projects let you group chats and knowledge together. Anthropic introduced Projects to organise curated knowledge and related chat activity in one place. 

A practical way to migrate a Custom GPT into a Claude Project:

1. Create a Project named after the assistant (e.g., “SEO Content Engine”).
2. Upload your key docs (brand voice, offers, product details, FAQs, style guide).
3. Add a “Project Instructions” file (a short version of your Assistant Spec).
4. Start a new chat inside the Project and paste:
   • The Assistant Spec
   • The latest Migration Brief
   • A sample task

Claude supports file uploads directly in chat (via the “+” / add files flow). 

Project File Kit: CLAUDE.md-Style Guidance

Even if you don’t use code repos, borrow the idea: keep one “source of truth” instruction doc per Project, and update it when you refine behavior.

Suggested files:

• PROJECT_BRIEF.md (what this project is)
• INSTRUCTIONS.md (tone + rules + output formats)
• EXAMPLES.md (great responses you want repeated)
• SOURCES.md (links and documents that matter)

This reduces “prompt drift” over time.

Step 7 – Move to Gemini: Export and Prepare Gemini Data

If you want to back up Gemini data (or prepare for a future switch), Google provides an official export route via Google Takeout for Gemini Apps data. The Gemini help page describes using Takeout and selecting Gemini-related data—including Gems data. 

This matters because a portable workflow is a safer workflow.

Step 8 – Rebuild Custom GPTs as Gemini Gems

Gemini Gems are custom assistants inside Gemini. Google’s help docs outline the basic creation flow: go to Gemini, open Explore Gems, choose New Gem, write instructions, preview, then save. 

Migration method (Custom GPT → Gem):

1. Open your Assistant Spec.
2. Paste the most important parts into the Gem instructions:
   • Role + mission
   • Rules
   • Output formats
   • Examples
3. Add your “startup prompt” (a short prompt you run first when starting a new task).
4. Save and test using a real scenario from your old ChatGPT threads.

Gem Instruction Checklist

• Keep rules short and prioritised
• Include examples (these matter a lot)
• Specify output formatting (“use a table,” “include a checklist,” etc.)
• Define what to do when missing info (“ask 3 questions then proceed with assumptions”)

Step 9 – Bring Conversation JSON Along (Even If It Can’t “Import”)

Chat export JSON is still valuable even when platforms don’t ingest it as “native chat history.”

You can:

• Store it as an archive
• Convert it into readable Markdown
• Upload snippets into a Project/Gem for reference
• Ask Claude/Gemini to summarise or extract reusable parts

OpenAI even documents a workaround approach for moving conversations between ChatGPT accounts by exporting JSON and uploading it into a new conversation (not a perfect recreation, but useful for portability). 

JSON → Markdown Conversion (Beginner-Friendly, No Coding Needed)

If you exported your ChatGPT data and you see a file like conversations.json, don’t panic. You don’t need to “understand JSON” to make it useful.

The goal is simple: turn your chat export into clean, readable notes you can upload into a Claude Project or keep as a searchable archive.

Option A: Use an Online Converter (Fastest)

If you only need to convert a few conversations quickly, use a converter tool:

• https://sitegpt.ai/tools/convert-json-to-markdown
• https://products.aspose.app/cells/conversion/json-to-markdown

How to do it (simple steps):

1. Open the converter tool in your browser
2. Upload your conversations.json file (or paste the content)
3. Convert to Markdown (.md)
4. Download the output and save it in a folder like:
   
   Archive/ChatGPT/Marketing/2026-03-02-content-ideas.md

Why Markdown?

It’s clean, readable, and works well when you upload files into Claude Projects or reference them later.

Privacy tip: If your chats include client info, passwords, or sensitive details, don’t upload the full file to a web tool. Either remove sensitive parts first, or use Option B below.

Option B: Skip Conversion and Create a “Migration Brief” Instead

For most people, this is the best path.

Instead of converting everything, take your top 10–20 most valuable conversations and create short “Migration Briefs” (500–900 words) that capture the important parts.

How:

1. Open the original ChatGPT conversation
2. Ask ChatGPT:
   
   “Summarise this into a Migration Brief I can paste into Claude/Gemini. Include: goal, context, constraints, key decisions, final outputs, reusable templates, and next steps.”
3. Save the result as a Markdown file:
   
   Migration/Briefs/2026-03-02-seo-workflow.md

This is usually more useful than raw JSON because it removes noise and keeps the “gold.”

Step 10 – Validate: Regression Tests for Your Most Important Workflows

Before you fully switch, run “regression tests”:

• Pick 5–10 tasks you do often (e.g., write a landing page, summarise a contract, generate a content outline).
• Use the same input in:
  • Your old setup (ChatGPT)
  • Your new setup (Claude Project or Gemini Gem)
• Compare outputs:
  • Accuracy
  • Tone match
  • Completeness
  • Format compliance

Then update your Assistant Spec with whatever is missing.

Step 11 – Secure Your Data: Redaction, Minimisation, Retention

Migration is also a privacy moment. Be intentional:

• Redact secrets: API keys, private client data, passwords.
• Minimise: don’t upload entire exports if you only need a few threads.
• Separate personal vs. business context into different Projects/Gems.
• Use summaries instead of raw logs when possible.

A clean Migration Brief is usually safer than dumping raw chat logs.

Step 12 – Maintain the Setup: A Simple Migration Maintenance Routine

To keep your setup future-proof:

• Monthly: export important new work (short summaries are fine)
• Quarterly: update Assistant Specs (version them like spec_v3.md)
• Anytime you change your business: update one “Master Context” brief
• Keep an archive: store exports and Markdown conversions offline

This turns “migration” from a stressful event into routine maintenance.

Conclusion: Your New “Multi-Model” Workflow

If you follow this guide, you won’t just switch assistants – you’ll upgrade your entire system. The secret is to stop thinking of your AI as a single chat app and start treating it like a portable knowledge workflow:

• Export your data officially
• Triage ruthlessly
• Summarise into Migration Briefs
• Rebuild assistants from clear specs
• Use Claude Projects or Gemini Gems for structure
• Keep your archives in Markdown for portability

That’s the real long-term win, and it makes How to migrate from ChatGPT to Claude or Gemini a clean, repeatable process instead of a one-time headache.

Are you on the lookout for a social media consultancy?

We’re OOB 👋  A a culture-first social marketing consultancy delivering for the world’s boldest brands, from TikTok to Ann Summers.

Big strategy. Killer creative. Creator campaigns with teeth. Built in London, fluent everywhere. Ready to make noise? 👉 hello@ourownbrand.co