ChatGPT Custom Instructions: The Complete Guide
Every time you start a ChatGPT conversation, it knows nothing about you. Custom instructions change that.
Here’s how to set them up for genuinely better results.
What Custom Instructions Are
Two text boxes that apply to every conversation:
“What would you like ChatGPT to know about you?”
- Your background, expertise, context
“How would you like ChatGPT to respond?”
- Format, tone, style preferences
Once set, every new conversation starts with this context.
Where to Find Them
On desktop: Click your name (bottom left) → Settings → Personalization → Custom instructions
On mobile: Tap menu → Settings → Custom instructions
The “About You” Section
This tells ChatGPT who you are. Better context = better responses.
Template Structure
Professional context:
[Your job/role]
[Your industry]
[Years of experience]
[Expertise areas]
Current focus:
[What you're working on]
[Goals you're pursuing]
Tools you use:
[Software, platforms, technologies]
Location/context:
[Relevant geographic or cultural context]
Example: Marketing Manager
I'm a digital marketing manager at a B2B SaaS company (50 employees).
10 years marketing experience, focused on content marketing and SEO.
Current focus: Building organic traffic and improving conversion rates.
We use HubSpot, Google Analytics, SEMrush, and Slack.
Target audience: IT decision-makers at mid-size companies.
Based in US, working with primarily North American clients.
Example: Software Developer
Senior full-stack developer, 8 years experience.
Primary stack: Python/Django, React, PostgreSQL, AWS.
Currently building a fintech SaaS product.
Team of 6 developers, I often mentor juniors.
Interested in clean architecture, testing, and DevOps practices.
Based in Europe, working for US company.
Example: Freelance Writer
Freelance B2B writer specializing in technology and finance.
5 years experience, clients include startups to Fortune 500.
I write blog posts, whitepapers, case studies, and web copy.
Focus on clear, concise writing without jargon.
Usually researching: fintech, AI, SaaS, productivity tools.
Work with mostly US and UK clients.
The “How to Respond” Section
This shapes ChatGPT’s output style. Be specific.
Template Structure
Tone and style:
[How should responses sound?]
Format preferences:
[Bullets vs. paragraphs? Headers? Length?]
What to include:
[Examples? Code? Steps?]
What to avoid:
[Things you don't want?]
Special instructions:
[Anything else relevant?]
Example: Direct and Practical
Be direct and concise. Skip unnecessary qualifiers.
Use bullet points for lists, not numbered unless sequence matters.
Don't explain things I likely already know based on my experience.
Give me the answer first, then the explanation if needed.
For code: Include comments only where non-obvious.
For writing: Avoid corporate jargon and buzzwords.
No need for disclaimers or excessive politeness.
Just give me what I asked for.
Example: Detailed and Educational
Explain concepts thoroughly, including the reasoning.
Use examples to illustrate points.
Structure with clear headers when appropriate.
Include pros/cons when discussing options.
Cite sources or methods when relevant.
If you're uncertain, say so rather than guessing.
I learn better with analogies - use them when helpful.
Example: Creative and Exploratory
Take creative risks. Don't just give safe answers.
Push boundaries and suggest unexpected approaches.
Ask clarifying questions rather than assuming.
Offer multiple perspectives on complex topics.
Use metaphors and vivid language.
Don't sanitize interesting ideas.
If something is genuinely debatable, present multiple views.
Real Custom Instructions That Work
For Business Professionals
About you:
Director of Operations at a logistics company, 200 employees.
15 years in supply chain management.
Currently implementing new WMS, improving warehouse efficiency.
Use SAP, Excel extensively, learning Power BI.
Need to communicate with both executives and floor staff.
How to respond:
Be concise and actionable. I have limited time.
Lead with recommendations, follow with reasoning.
When discussing data/numbers, be precise.
Adjust complexity based on stated audience.
Use industry terminology - I know logistics.
Provide templates or examples when applicable.
For Students/Researchers
About you:
PhD candidate in cognitive psychology.
Research focus: decision-making under uncertainty.
Strong stats background (R, Python for analysis).
Publishing in peer-reviewed journals.
Also TA for undergrad methods courses.
How to respond:
Be academically rigorous. I value precision.
Cite relevant research when applicable.
For stats questions, show the reasoning, not just answers.
Distinguish between established findings and speculation.
When I'm prepping teaching materials, adjust for undergrad level.
Challenge my assumptions when appropriate.
For Creatives
About you:
Freelance UX/UI designer, 6 years experience.
Specialize in mobile apps and SaaS dashboards.
Use Figma, conduct user research, work closely with developers.
Current clients: fintech and healthtech startups.
Also run a design newsletter.
How to respond:
Think visually. Describe layouts, flows, interactions.
Reference design patterns and best practices.
Consider accessibility and usability, not just aesthetics.
When reviewing my work, be critically honest.
For newsletter content: conversational, not academic.
Include practical examples from real products when relevant.
Common Mistakes to Avoid
Too Vague
Bad: “I work in tech and want good answers”
Good: “Senior DevOps engineer at a fintech startup. I use Kubernetes, Terraform, AWS. Focus on reliability and automation.”
Too Long
Custom instructions have character limits. Prioritize what matters most. You don’t need your life story.
Static Thinking
Update your instructions as your focus changes. What you needed six months ago may not be relevant now.
Conflicting Instructions
“Be concise” + “Explain everything thoroughly” creates confusion. Pick a primary mode.
Specialized Setups
For Coding
About:
Languages: Python, TypeScript, Go
Frameworks: FastAPI, React, Next.js
Patterns: Clean architecture, TDD
Environment: macOS, VS Code, Docker
Response:
Code examples should be complete and runnable.
Use type hints in Python, TypeScript types.
Include error handling in examples.
Prefer modern syntax and patterns.
Comment only non-obvious parts.
For Writing
About:
Professional writer: tech, business, productivity.
Style: Clear, direct, no fluff.
Publications: [Your outlets]
Typical formats: 1500-2000 word articles.
Response:
Write in active voice.
Lead with value, not introductions.
No buzzwords: leverage, synergy, etc.
Structure with scannable headers.
End with actionable takeaways.
For Research
About:
Market researcher for consumer products.
Methods: Surveys, focus groups, data analysis.
Tools: Qualtrics, SPSS, Tableau.
Report to CMO and product teams.
Response:
Distinguish data from interpretation.
Include methodological considerations.
Translate findings to business implications.
Use appropriate hedging for uncertain conclusions.
Suggest follow-up research when warranted.
Testing and Refining
After Setting Up
- Start a fresh conversation
- Ask something you commonly ask
- Note what’s different/better
- Identify remaining gaps
- Adjust instructions
Ongoing Refinement
- When ChatGPT misses context, add it to instructions
- When responses are too long/short, adjust preferences
- When focus changes, update current projects
The Bottom Line
Good custom instructions take ChatGPT from “generic AI” to “AI that knows you.”
Start with:
- Your professional context (role, expertise)
- What you’re currently working on
- How you like information presented
Refine based on:
- Repeated context you find yourself adding
- Format issues that keep coming up
- Areas where ChatGPT consistently misses
10 minutes of setup saves hours of repeated context. Update them quarterly as your work evolves.
Frequently Asked Questions
Custom instructions tell ChatGPT about you and how you want it to respond. They apply to every conversation, so you don't repeat the same context each time.
Significantly. Generic ChatGPT doesn't know your expertise, preferences, or communication style. Custom instructions add that context, making responses more relevant.
Not natively - you get one set. But you can create custom GPTs with different instructions for different use cases, or manually update instructions when your needs change.