ChatGPT Memory and Custom Instructions Guide

Master ChatGPT memory features and custom instructions to get better, more personalized responses.

Mastering ChatGPT Personalization

ChatGPT's memory and custom instructions features transform it from a generic assistant into a personalized AI that understands your context, preferences, and needs. Here's how to use them effectively.

Understanding Memory

What Memory Does:

  • Remembers facts you share across conversations
  • Builds context about your work and preferences
  • Persists between chat sessions
  • Can be viewed and managed
  • What Memory Doesn't Do:

  • Remember everything automatically
  • Store sensitive information (it filters these out)
  • Work across different accounts
  • Remember if you're in a temporary chat
  • Managing Memory

    To View Memory: Settings → Personalization → Memory → Manage

    To Add to Memory: Simply tell ChatGPT:

  • "Remember that I'm a software developer using Python"
  • "Remember my company is called TechStart"
  • "Remember I prefer concise responses"
  • To Remove from Memory:

  • Click "Manage" and delete specific memories
  • Say "Forget that I told you about X"
  • Clear all memory in settings
  • Custom Instructions Setup

    Custom Instructions are more powerful than memory. They're applied to EVERY conversation.

    Accessing Custom Instructions: Settings → Personalization → Custom Instructions

    Two Sections:

  • "What would you like ChatGPT to know about you?"
  • - Your role/profession - Your expertise level - Your projects/context - Your preferences

  • "How would you like ChatGPT to respond?"
  • - Response format - Tone and style - What to include/exclude - Technical level

    Custom Instructions Templates

    For Developers:

    
    ABOUT ME:
    
  • Senior full-stack developer
  • Primary: Python, TypeScript, React
  • Using VS Code on Mac
  • Building SaaS products
  • RESPONSE STYLE:

  • Code first, explain after
  • Use modern best practices
  • Include error handling
  • No basic explanations
  • Suggest optimizations
  • For Writers:

    
    ABOUT ME:
    
  • Content writer for tech companies
  • Writing blog posts and documentation
  • Target audience: business professionals
  • Focus on clarity and engagement
  • RESPONSE STYLE:

  • Clear, conversational tone
  • Use active voice
  • Short paragraphs
  • Include examples
  • Avoid jargon unless necessary
  • For Business:

    
    ABOUT ME:
    
  • Marketing manager at B2B SaaS company
  • Team of 5, $500K annual budget
  • Focus on content marketing and SEO
  • Using HubSpot, Semrush, Google Analytics
  • RESPONSE STYLE:

  • Be direct and actionable
  • Include metrics where relevant
  • Consider resource constraints
  • Suggest priorities
  • Business-appropriate language
  • Advanced Tips

    1. Be Specific Bad: "I like coding" Good: "I'm a Python developer with 5 years experience, working on Django web apps"

    2. Set Boundaries

  • "Never start responses with 'Great question!'"
  • "Don't use emojis"
  • "Skip pleasantries, get to the point"
  • 3. Define Output Format

  • "Use markdown formatting"
  • "Include code blocks with syntax highlighting"
  • "Structure responses with headers"
  • "End with action items when appropriate"
  • 4. Expertise Level

  • "Assume I understand programming basics"
  • "Explain like I'm new to marketing"
  • "Use technical terms, I'll ask if confused"
  • Memory + Custom Instructions Strategy

    Use Memory For:

  • Specific project details
  • Current goals and deadlines
  • Team member names
  • Recent decisions made
  • Use Custom Instructions For:

  • Your role and expertise
  • Response preferences
  • Consistent formatting
  • Things that apply to all chats
  • Example Workflow

    Initial Setup:

  • Fill out custom instructions with your role and preferences
  • Start using ChatGPT normally
  • Explicitly add important context to memory
  • Periodically review and clean up memory
  • Ongoing:

  • "Remember I decided to use PostgreSQL for this project"
  • "Forget the old deadline, new one is March 15"
  • "Remember my manager's name is Sarah"
  • Common Mistakes

  • Too vague - "I like short responses" → "Keep responses under 200 words unless I ask for detail"
  • Too restrictive - Don't limit so much that ChatGPT can't help
  • Not updating - Review quarterly as your needs change
  • Conflicting instructions - Make sure memory and custom instructions align
  • Privacy Considerations

  • Memory data is used to improve models (unless you opt out)
  • Don't store sensitive info (passwords, SSN, etc.)
  • ChatGPT filters out some sensitive data automatically
  • Use temporary chats for sensitive queries
  • You can disable memory entirely in settings
  • Testing Your Setup

    After configuring, test with:

  • "What do you know about me?"
  • Ask a question in your field
  • Check if response matches your preferred style
  • Adjust as needed
  • Platform Differences

  • Web/iOS/Android: Full memory and custom instructions
  • API: No memory (you manage context)
  • Teams/Enterprise: Memory may be limited by admin

Good personalization means better responses with less prompting. Take 15 minutes to set this up properly and you'll save hours over time.

Share this article: