Custom Instructions: Your Professional Profile

Custom Instructions is one of the most underused ChatGPT features. It lets you set a persistent context that applies to every conversation — so you stop re-explaining who you are, what you do, and how you want responses formatted every single time you start a new chat.

What Custom Instructions Does

Without custom instructions, every conversation starts from scratch. ChatGPT knows nothing about you, your industry, your role, your preferences, or your workflow. You waste the first few messages establishing context.

With custom instructions, you provide a one-time setup that applies to every conversation. ChatGPT knows you're a product manager at a SaaS company, that you prefer concise responses, that you want bullet points rather than paragraphs, and that your primary audience is non-technical stakeholders — without you saying any of it.

Accessing Custom Instructions

ChatGPT → Your profile (bottom left) → Customize ChatGPT
Or: Settings → Personalization → Custom instructions

Two text fields:

"What would you like ChatGPT to know about you?" — Your background, role, context
"How would you like ChatGPT to respond?" — Format preferences, tone, style

Writing Effective Custom Instructions

Field 1: What to know about you

Be specific and professional. Think of it as briefing a very capable new assistant:

I'm a senior product manager at a B2B SaaS company (50-person startup, 
Series A). My work spans product strategy, user research, writing PRDs, 
stakeholder communication, and prioritization decisions.

My industry: HR tech (talent acquisition and employee onboarding software)
Primary users: HR managers and recruiters at mid-market companies
My technical level: I can read code but don't write it

Key context:
- My team is 4 engineers, 1 designer, 1 data analyst
- We run 2-week sprints
- Decision makers I communicate with: CEO, VP Engineering, VP Sales
- Primary KPIs: user activation, feature adoption, NPS

Field 2: How you want responses

Response style:
- Be direct and concise — I prefer dense, useful content over padding
- Lead with the most important information
- Use bullet points and numbered lists for action items and options
- Keep explanations short unless I ask to "explain fully"
- If you're unsure about something, say so clearly rather than hedging

Formatting:
- Use markdown when it adds clarity (headers, tables, code blocks)
- For decisions/tradeoffs: use a table or structured comparison
- For action items: numbered list with clear ownership
- Max 3-4 paragraphs unless I ask for more depth

Things to avoid:
- Don't use filler phrases like "Great question!" or "Certainly!"
- Don't add disclaimers unless they're essential
- Don't repeat back what I just said before answering

Custom Instructions by Role

Developer

Who I am:
Full-stack developer, 5 years experience. Primary stack: TypeScript, React, 
Next.js, PostgreSQL, Prisma, deployed on Vercel. Working at a 20-person startup 
as the only frontend developer.

How to respond:
- Show code before explaining it
- Use TypeScript (strict mode) in all code examples
- Prefer functional components and hooks over class components
- Use modern ESM syntax, not CommonJS
- When there's a simpler way to do something, mention it
- Don't add error handling I didn't ask for
- Assume I know basic concepts; skip beginner explanations

Marketing Manager

Who I am:
Content marketing manager at a B2C e-commerce company. I write blog posts, 
email campaigns, social media content, and product copy. Primary audiences: 
women 25-45 interested in sustainable home goods.

Brand voice: Warm, friendly, authentic. Not pushy. We're anti-corporate.

How to respond:
- Match the tone: conversational but not casual, clear not clever
- No jargon, no corporate speak
- When writing copy: lead with the reader's benefit, not the product feature
- For email subject lines: give me 5 options with different emotional angles
- Keep headlines under 8 words when possible

Business Analyst / Data Professional

Who I am:
Senior business analyst at a financial services firm. I analyze data using SQL 
and Python (pandas), create dashboards in Tableau, and present insights to 
C-suite executives.

How to respond:
- For data questions: show SQL/Python first, then explain
- For presentations: executive-friendly language, metrics-first
- Quantify everything where possible
- Flag assumptions clearly
- Use tables for data comparisons

Testing Your Custom Instructions

After saving, start a new conversation and ask something work-related. Check:

Does it know your role without you mentioning it?
Is the format what you specified?
Is the tone right?

Refine over time. If you notice ChatGPT doing something that annoys you in multiple conversations — add a rule to your custom instructions.

Tips for Maximum Value

Be specific, not generic. "I'm a developer" is weak. "I'm a backend Python developer specializing in FastAPI and data pipelines at a fintech company" gives ChatGPT real context to work with.

Specify what NOT to do. "Don't add disclaimers" and "Don't use filler phrases" are often more valuable than positive instructions because ChatGPT has strong defaults toward cautious, padded responses.

Update as your role changes. If you change companies, switch teams, or start a new project, update your instructions.

Create separate GPTs for separate contexts. Custom instructions are one profile. If you have very different use cases (work vs. personal learning vs. a side project), create a custom GPT for each context with tailored instructions.

Next lesson: Understanding GPT-4o vs o1 vs o3 models — which model to use and when.