If you’ve been using AI as a tool for any length of time, you probably started with ChatGPT. It was the first model to break into the mainstream and bring generative AI capabilities to the masses. For the first couple of years after its launch, it felt like the only game in town.

Today, the AI landscape has changed; other LLMs like Google Gemini, Claude, and Grok can match and in some cases exceed ChatGPT’s capabilities. Each brings different features, different strengths, and a noticeably different “personality.” At Puritas, we became interested in Gemini’s capabilities because it has a massive context window, which means it can work with very large chunks of text, positioning it as a better tool for working with things like contracts, wills, and other legal documents.

At some point, loyal ChatGPT users like us might start to think: Should I switch? Or at least, should I try another model?

Switching AI tools, however, isn’t quite like switching to a different program. Moving from Word to Google Docs is a matter of learning some new menus and converting some files. When you move to a new AI, you leave all your past instructions and conversations behind. You can’t just transfer some files; you also need to transfer context: the accumulated understanding the AI has built about how you think, what you’re working on, and how you like things done.

What to Take With You

Just switching AI assistants without bringing over the context the old assistant has accumulated will prevent you from making a fair comparison. You need to bring the new AI up to speed on your projects, preferences, and general nuances that your current AI already knows.

Here’s a quick breakdown of what you will need to properly introduce yourself to a new AI.

Personal Context

This includes your goals, the kinds of projects you work on, your professional role, and your general expectations. Over time, good AI tools learn whether you’re a lawyer, a writer, a business owner, or an accountant, and they tailor their answers accordingly.

Persona and Style

This is typically the information you put in your “custom instructions”. Maybe you prefer concise answers. Maybe you want bullet points only. Maybe you dislike motivational fluff and just want the facts. These preferences are part of your working relationship with an AI and are worth transferring.

Ongoing Project Details

Long-running projects build up an invisible state: prior decisions, agreed terminology, abandoned approaches, and constraints. Without this history, a new AI will happily suggest things you already ruled out.

How to Pack Up Your Context and Hit the Road

There are a couple of services that offer one-click “sync my AI brain” buttons. But we have found that a manual process, best thought of as briefing a new assistant, works just fine and actually can be enlightening to both you and your new AI.

1. Extract and Summarize

This can be done at a general level or within specific chats. Just ask your current AI to summarize what it knows about you or a specific project. Here is an example:

“Summarize my recurring preferences, ongoing projects, and decision patterns in structured topic clusters.”

This turns your AI’s implicit knowledge into something portable.

2. Move Your Context

You can just dump some of the portable context you have gathered into a chat on your new AI and hit the ground running. However, some things you may want to persist throughout all your chats, so you will want to make them part of your profile. Each platform has its own way of storing this type of persistent context:

• Gemini: You can add key facts to the Saved info section so they persist across chats.

• ChatGPT: Uses Custom Instructions to define your background, goals, and preferences. Check the Projects article elsewhere in this email.

• Claude: Stores information in Memory, or you can use Projects to upload reference documents and long-term context.

This approach reduces your need to train or repeat instructions each time you start a chat.

3. Quiz Your New AI to Confirm what it Now Knows

After transferring context, ask the new AI to paraphrase what it has learned. This is an important calibration step that surfaces misunderstandings early, before they compound.

Bottom line: Without context your new AI won’t know how you like your coffee, how you format emails, or why Project B was shelved. Transferring context lets you make a fair comparison of the new AI and get productive faster.


That said, context isn’t always an asset. Over time, outdated assumptions, abandoned projects, and old preferences can weigh an AI down just as easily as they can help it. In some cases, starting fresh—produces clearer thinking and better results. 

The key is being intentional: carry over what still serves you, and leave behind what doesn’t.