How to Set Up Claude AI Personal Preferences (And Why It's the First Thing You Should Do)
How to Set Up Claude AI Personal Preferences (And Why It's the First Thing You Should Do)
Claude AI Personal Preferences are persistent instructions that tell Claude who you are, how you work, and how to respond to you. You set them once in Settings > Profile, and they apply to every conversation automatically. They are the single most impactful thing you can configure in Claude, and they're available on every plan including the free tier.
Most people either don't know this setting exists or stare at the empty text box and don't know what to write. The result: every conversation starts from scratch, Claude doesn't know your context, and the answers feel generic.
This guide covers exactly what to include, common mistakes to avoid, and how to build yours in about 5 minutes using our free tool.
What Are Claude Personal Preferences?
Personal Preferences is a text field in Claude's settings where you tell Claude about yourself: your role, your communication style, how you like to work, and what you want Claude to do (or not do) in every conversation.
Once saved, Claude reads your preferences at the start of every new conversation automatically. You set them once. They apply everywhere.
You'll find them in Settings > Profile > "What personal preferences should Claude consider in responses?" on claude.ai. The feature is available on every plan, including the free tier.
Why Do They Matter?
Without preferences, Claude defaults to being polite, thorough, and generic. That's fine for a one-off question. It's frustrating when you use Claude regularly.
Here's what changes when you set up good preferences:
Claude stops giving generic answers. Instead of treating every question like it's from a stranger, Claude knows your context. A founder asking about hiring gets a different answer than a student researching the same topic.
You stop repeating yourself. No more typing "keep it brief" or "use bullet points" at the start of every conversation. Your preferences handle that automatically.
Claude matches your tone. If you want casual and direct, you get casual and direct. If you want formal and thorough, you get that instead. Every time, without asking.
Claude pushes back when you want it to. Default Claude is agreeable. It validates your ideas, hedges its answers, and avoids disagreement. With the right preferences, you can tell Claude to challenge your thinking and flag when your approach has flaws.
Claude knows what to skip. If you're an expert in finance, Claude won't explain basic concepts. If you're learning a new skill, it will break things down step by step.
What Should You Include?
This is where most people get stuck. They stare at an empty text box, type something like "be concise," and move on. That barely scratches the surface.
Good preferences cover six areas:
1. Who You Are
Your role, industry, and what you primarily use Claude for. This shapes how Claude calibrates its answers.
A marketing manager gets different depth and terminology than a software engineer, even for the same question. Without this context, Claude guesses. And it usually guesses too broad.
Example:
- I'm a Founder/CEO in retail and e-commerce, based in the UAE
- I use Claude daily for writing, brainstorming, research, and building internal tools
2. Communication Style
How you want Claude to talk to you. This includes response length, structure, tone, and language.
Most people don't realise how much control they have here. You can tell Claude to default to bullet points, use British English, keep answers under a certain length, or match a specific formality level for different contexts (formal for work emails, casual for brainstorming).
Example:
- Keep responses short and direct
- Default to bullet points and lists
- Use British English spelling and phrasing
- Casual tone, like a smart friend, not a corporate assistant
3. How You Work
This is where preferences become genuinely powerful. These are the behaviours most people don't know they can control.
Should Claude give you a recommendation or just present options? When you ask something vague, should it clarify first or take its best shot? Should it push back on your ideas?
Example:
- Present options with a clear recommendation when I'm deciding
- If something's ambiguous, ask me one clarifying question first
- Push back on my ideas if you think I'm wrong; I welcome it
- Give me one step at a time on complex tasks; I'll ask for the next step
4. Your Expertise
This creates a personalised depth calibration. Claude skips the basics on topics you know and explains more on topics you're learning.
The contrast between "expert" and "learning" topics is what makes this work. Without it, Claude either patronises you on your strengths or overwhelms you on new subjects.
Example:
- Expert in operations and finance: skip fundamentals, go straight to application
- Learning data analytics and HR: explain concepts before jumping in
- I write some code but I'm not a developer: keep technical language plain
5. Things Claude Should Always Do
Positive behaviours you want Claude to maintain in every conversation. These are easy to overlook but high impact.
Example:
- Cite sources when possible
- Warn me before anything irreversible
- Flag when you're guessing vs. confident
- Search the web before giving product or pricing advice
6. Things Claude Should Never Do
This is the section that makes the biggest difference. Default Claude has behaviours that drive regular users crazy. You can switch them off.
Example:
- Never open with "Great question!" or similar praise
- Don't apologise excessively
- Don't repeat my question back before answering
- Don't hedge everything with "it depends"
- Don't use filler phrases ("Absolutely!", "Of course!", "Certainly!")
- Don't give me walls of text when I asked something simple
Most people have no idea these behaviours are controllable. Once you switch them off, the difference is immediate.
Common Mistakes
Too vague. "Be helpful and concise" tells Claude almost nothing. Every bullet should be specific enough that it changes how Claude actually behaves. "Start with the answer, then explain only if I ask" is specific. "Be concise" is not.
Too long. Your preferences consume tokens in every conversation. If you write 2,000 words of preferences, you're using up context window before you've even asked a question. Aim for 300 to 500 words. That's the sweet spot between comprehensive and efficient.
Only positive instructions. Telling Claude what NOT to do is often more impactful than telling it what to do. Claude's default behaviours are mostly fine. It's the specific annoying ones you want to eliminate.
Set and forget. Good preferences evolve. Every time you notice Claude doing something that frustrates you, add a line to your preferences to fix it. The best preferences are the ones that have been refined over weeks and months.
How Preferences Fit Into Claude's Personalisation Layers
Personal Preferences are one of three ways to customise Claude. Understanding how they work together helps you put the right instructions in the right place.
Personal Preferences apply to every conversation, across all projects. This is where your general identity, communication style, and behavioural rules live. Think of it as "who I am and how I work."
Project Instructions apply only within a specific Claude Project. If you're working on a web app, you might set project instructions like "This is a Next.js project using TypeScript and Supabase." These are scoped and specific.
Styles control how Claude formats and delivers responses. You can switch between styles mid-conversation. Think of these as "how Claude writes" rather than "how Claude thinks."
Your Personal Preferences are the foundation. Get these right first. Everything else builds on top.
Build Your Preferences in 5 Minutes
Writing good preferences from a blank page is hard. Knowing what to include, how specific to be, and how to stay within the right word count takes trial and error. I rewrote mine probably ten times before finding the right balance.
That's why we built a free tool that does it for you.
You answer 6 pages of multiple choice questions about your role, how you think, how you communicate, and what annoys you about AI. Then AI generates personalised preferences you can paste straight into Claude's settings.
The output includes smart inferences you wouldn't think to specify yourself, like applying your local legal context automatically or leading with executive summaries when you ask for documents.
Takes about 5 minutes. The result is preferences that are specific, well structured, and optimised to stay under 500 words so they don't eat into your context window.
What's Next?
Personal Preferences are step one. Once they're set up, the next level is Projects: dedicated workspaces where Claude maintains context, follows project-specific instructions, and references your documents.
We'll cover that in a future guide. But none of it matters if the basics aren't set up first. Start with your preferences.