Tell Claude Who You Are

Complete What AI Can't Know (And How to Check) first; this lesson builds on knowing what AI can and cannot reliably infer about you without being told.

Every new conversation with Claude starts with a blank slate. It has no memory of your last chat, no idea what job you do, and no way to know how you prefer responses unless you tell it. Most people re-explain themselves every single time: 'I am a project manager, I work in healthcare, please keep your answers short.' That is wasted effort repeated in every conversation, for as long as you use the tool.

Every major AI tool has a one-time setup that solves this permanently. In Claude, it is called a Profile. In ChatGPT, it is called Custom Instructions. In Gemini, it is called Gems. You write a short description of yourself once, and the AI reads it at the start of every conversation from that point on. You never have to re-explain yourself again.

Here is the before and after: Before: no profile, every chat starts with 'I am a marketing manager at a mid-size company.' After: a profile that reads, 'I am a marketing manager at a mid-size consumer goods company. I work mostly on campaigns and vendor relationships, not code or data. Keep responses conversational and under 200 words unless I ask for more. Avoid jargon and technical terms.' That 40-word profile changed the quality of every single conversation that followed.

Keep it short. Three to five sentences covers role, current focus, and preferences. You can update it any time. Ask Claude to help you write it.

Paste this into Claude:

I want to set up my profile so you always know who I am. Here's what's relevant: I am [job/role]. I'm working on [current focus]. I prefer responses that are [short and direct / detailed with examples / structured with bullet points]. I don't have a technical background, so [avoid code / explain jargon / assume I'm using web tools]. Save this as my profile.

What good looks like:

• You found the profile settings in your AI tool (in ChatGPT: click your name, then 'Customize ChatGPT'; in Claude: click your initials, then 'Profile') and saved a short description of yourself
• You started a new conversation and the AI responded using your profile without you repeating yourself
• The response felt tailored to you rather than generic
• Your profile is five sentences or shorter and covers role, current focus, and at least one preference about how you like responses
• You tested it in a fresh conversation and could tell the difference from a no-profile response

What a good response looks like:

Here's a profile you can paste directly into Claude or ChatGPT Custom Instructions:

'I am a marketing manager at a mid-size consumer goods company. I'm currently focused on vendor contract renewals and preparing a campaign brief for Q1. Keep your responses conversational and under 200 words unless I ask for more. Avoid jargon and technical terms. When I ask for emails or messages, default to a warm, professional tone, not corporate-formal.'

That's five sentences. It covers your role, your current focus, and three specific preferences about how you want responses formatted. When you save this and start a new conversation, Claude will respond to your first message already calibrated to those preferences, without you having to explain any of it. Want me to adjust the tone, shorten it further, or add anything specific about your work style?

Paste this into Claude:

I have a profile set up, but I want to make sure it is actually changing the responses I get. Start a new conversation with my profile active. I am going to give you a task I do regularly: [describe a real recurring task, like 'help me write a weekly update email to my team' or 'help me prep talking points for a client call']. Give me a response based on my profile, then tell me: what specific things in my profile shaped your answer? What would you have done differently without it?

What good looks like:

• The AI explicitly named at least one thing in your profile that shaped its response
• The response was noticeably more tailored than what a profile-less version would produce
• You identified at least one preference in your profile you want to add or change based on this test
• You updated the profile with the improvement before ending the session

What a good response looks like:

Here are the talking points for your Friday client check-in with Meridian Packaging:

1. Q3 campaign results: impressions up 18%, click-through rate down slightly at 2.1% vs 2.4% target. Worth acknowledging directly.
2. Timeline update: creative assets delivered on schedule. Next milestone is media buy confirmation by Oct 15.
3. One ask: we need their brand refresh guidelines before we finalize the November assets.

Based on your profile, I kept this under 150 words, used plain language, and skipped the standard 'let me know if you have questions' filler. Without your profile, I would have written this as a longer narrative paragraph with a formal sign-off. The profile told me you prefer direct and short, so I structured it as a scannable list instead. One thing I noticed was missing: you did not mention whether you prefer bullet points or a numbered list for talking points. Adding that to your profile would tighten future responses.