When you first start using Claude, it doesn't know anything about you. That's by design. Claude doesn't assume anything until you tell it or until it learns from your conversations over time.

But "over time" can feel slow when you're used to another tool that already knows your preferences. So here's how to speed that up, whether you're building from zero or bringing context from somewhere else.

How Claude's memory works

Claude's memory system is different from ChatGPT's or Gemini's. Instead of saving a list of individual facts, Claude synthesizes what it learns into a profile that it updates in daily cycles. You can think of it like this: ChatGPT remembers "user prefers TypeScript" as a discrete entry. Claude builds an understanding that you're a developer who works in TypeScript, which it then uses to shape its responses more broadly.

You can view what Claude has learned about you in Settings > Privacy > Memory preferences. You can also edit memory preferences there, and Claude will incorporate your edits in its next daily cycle.

The key thing to know: Claude's memory gets better the more you use it, but you don't have to wait. You can kick-start it.

Option 1: Starting fresh

If you're new to AI assistants (or just want a clean start), the fastest way to set up Claude is to tell it about yourself in your first conversation. Don't overthink this. Just cover the basics.

Here's a solid starting prompt:

Setup prompt
Here's some context about me so you can tailor your responses: - Name: [your name] - Role: [what you do] - I prefer [concise/detailed] responses - When writing, match a [casual/professional/technical] tone - Tools I use regularly: [list them] - I'm currently working on: [projects or goals] - Pet peeves: [things you don't want, like "don't start with 'Great question!'"] Please remember all of this.

Claude will incorporate this into its memory. You don't need to repeat it in every conversation. From this point on, you can just correct Claude when something is off ("shorter answers please" or "use Python, not JavaScript") and it'll update its understanding.

You can also go to Settings and edit your memory preferences directly. Add things like "I prefer bullet points for technical explanations" or "I work in the healthcare industry." Claude picks these up in its next synthesis cycle.

Option 2: Importing from another tool

If you've been using ChatGPT, Gemini, or Copilot, you don't need to manually retype everything. You can export your existing memory and import it directly into Claude.

The process is dead simple:

1. Use our export prompt in your current tool to pull out everything it knows about you. We have specific prompts for ChatGPT, Gemini, and Copilot.

2. Copy the response.

3. Go to claude.ai/import-memory (or Settings > Privacy > Memory preferences > Start import) and paste it in.

That's it. Claude processes the import and starts using your preferences right away, with full integration happening within 24 hours during its next synthesis cycle.

Quick tip: paste the export into the AI Memory Editor before importing it. Remove anything outdated (old job titles, finished projects) and correct anything that's wrong. Garbage in, garbage out.

Fine-tuning after setup

Whether you started fresh or imported, you'll want to fine-tune over the first week or so. Here's what that looks like in practice:

Correct in real time. If Claude gives you a response that's too long, tell it. If it uses the wrong framework, correct it. These corrections feed into memory and Claude adjusts.

Check what it learned. After a few days, ask Claude "What do you know about me?" and review the response. If anything is missing or wrong, tell it to update.

Use memory preferences. For things you want Claude to always do (or never do), add them in Settings > Privacy > Memory preferences. These act as persistent instructions that Claude applies across all your conversations.

Using memory with Projects

If you use Claude for different kinds of work, Projects are worth setting up. Each Project gets its own memory space, so your work context doesn't bleed into personal conversations and vice versa.

For example, you might have a Project for coding work (where Claude knows your stack and coding style) and use general conversations for everything else (where Claude knows your personal preferences). The memory stays separate, and Claude adapts accordingly.

You can import memory into specific Projects too, which is handy if you're migrating context from a specific use case on another platform.

The bottom line: Claude's memory is flexible enough to work however you need it to. Whether you spend two minutes importing your old preferences or build it up naturally over time, the goal is the same. Get Claude to the point where it just gets you, without you having to explain yourself every time.