chatgpt-personalisation-overview
ITIAN ChatGPT Academy
Technology Simplified — Solutions That Work
Personalisation Overview
Learn how ChatGPT’s personalisation layers work together, decide where each preference belongs, and create a setup that is useful, consistent and privacy-aware.
Learning Outcomes
By the end of this lesson, you should be able to:
Recognise the Layers
Explain the roles of prompts, personality, characteristics, Custom Instructions, memory, Projects and Temporary Chat.
Choose the Right Scope
Put a preference in the smallest layer that matches where and how long it should apply.
Resolve Conflicts
Find repeated or contradictory instructions that can produce inconsistent responses.
Protect Your Privacy
Personalise without treating ChatGPT as a secure store for passwords, secrets or highly sensitive information.
The guiding principle: match the scope
A one-off request belongs in the current prompt. A general communication preference may belong in Personality, Characteristics or Custom Instructions. Rules and reference material for one continuing body of work belong in its Project. Useful personal details may be remembered only when you choose to use Memory.
Use the smallest personalisation layer that reliably covers the need.
Your Personalisation Map
These seven layers cover most situations. Availability and labels may vary by plan, device, region and product rollout.
Current Prompt
A temporary instruction for this answer or conversation.
Personality
A broad communication style; it does not add capability or change safety rules.
Characteristics
Fine-tunes qualities such as warmth, brevity, formatting and emoji where available.
Custom Instructions
Explicit recurring guidance about you and how you want responses written.
Memory
Saved details and, where enabled, useful context from past chats.
Project Context
Instructions, files and chats scoped to one continuing project.
Temporary Chat
A clean conversation that does not use or create memories; still avoid secrets.
What Each Layer Should Do
| Layer | Best for | Not the right tool for | Example |
|---|---|---|---|
| Current prompt | One task, answer or temporary exception | Preferences you want in every future chat | “Explain this without technical jargon.” |
| Personality | Overall tone and conversational feel | Improving factual accuracy or granting new abilities | Select a friendly communication style. |
| Characteristics | Adjusting style dimensions where the feature is available | Detailed workflows or project rules | More concise, warmer, fewer headings. |
| Custom Instructions | Stable, explicit preferences across chats | Passwords, access keys or one-project requirements | “Use New Zealand English and teach step by step.” |
| Memory | Useful details you want ChatGPT to remember | A guaranteed factual record or confidential database | Your preferred name or recurring learning goal. |
| Project | One body of work with its own instructions, chats and files | Preferences that should apply everywhere | ITIAN page template rules and academy files. |
| Temporary Chat | A clean conversation without memory reference or creation | Long-running work that needs continuity | A one-off discussion you do not want remembered. |
How the Layers Work Together
Direct Context Matters
Your current request and the content being discussed strongly shape the answer. A specific request can make the response differ from your general style setting.
Projects Create a Boundary
Project instructions apply inside that Project and can override global Custom Instructions. Keep project-only standards there.
Style Is Not Capability
Personality and Characteristics change presentation, not the model’s knowledge, safety rules or access to tools.
Memory Adds Context
Memory can make conversations more relevant, but important facts should still be checked and sensitive information minimised.
When answers feel inconsistent
- Check the current prompt for a competing instruction.
- Review Project instructions if you are inside a Project.
- Compare Custom Instructions, Personality and Characteristics.
- Ask ChatGPT what relevant memories it is using, then review Memory settings.
- Remove duplicates, simplify wording and test again in a fresh chat.
Interactive Personalisation Layer Finder
Describe a non-sensitive preference. This local tool recommends where it should live; it does not send or save your entries.
Three Practical Examples
Across Most Chats
Need: “Use New Zealand English and teach step by step.”
Best home: Custom Instructions. It is stable, explicit and broadly useful.
One ITIAN Project
Need: “Use the ITIAN navy, green and gold HTML template.”
Best home: Project instructions and reference files, because it belongs to one body of work.
One Formal Email
Need: “For this answer, use a formal and concise tone.”
Best home: The current prompt. Do not change a global setting for one task.
A Useful Personal Detail
Need: “Remember that I am learning website accessibility.”
Best home: Memory, if enabled and if you are comfortable saving the detail.
Test Your Setup
Run the same four safe tests after changing a setting. Compare tone, structure and relevance—not just whether the answer sounds pleasant.
| Test | Example request | Observe |
|---|---|---|
| Explanation | “Explain browser cookies to a beginner.” | Warmth, complexity, examples and jargon. |
| Writing | “Rewrite this short notice professionally.” | Language conventions, tone and length. |
| Checklist | “Turn this task into an accessible checklist.” | Formatting, headings and action clarity. |
| Planning | “Help me plan a small website update.” | Whether project context appears only where intended. |
Common Mistakes to Avoid
Putting Everything Global
Project-only rules in Custom Instructions can affect unrelated conversations. Move them into the relevant Project.
Expecting Style to Add Accuracy
A confident or professional personality does not make information more correct. Verify important claims.
Confusing Off with Deleted
Turning Memory off does not by itself delete saved memories. Review and delete them separately when needed.
Oversharing
Never enter passwords, authentication codes, private keys or information you are not authorised to share. Minimise sensitive details.
Recommended: 16:9 • Captioned • Desktop and mobile views
Future Guided Demonstration
This video will show how to configure, test and simplify a personalisation setup.
- Choose a personality and characteristics.
- Write focused Custom Instructions.
- Review Memory and privacy controls.
- Move project rules into a Project.
Practical Activity: Personalisation Audit
Review your own setup without entering any private content into this page.
0 of 10 completed — begin with the scope of your preferences.
Knowledge Check
Answer all five questions, then check your result.
Official Sources and Further Reading
Reviewed against current first-party OpenAI guidance on 13 July 2026.
OpenAI guidance used in this lesson
- Customising your ChatGPT personality — style, interaction with instructions and memory.
- Characteristics in ChatGPT — fine-tuning response qualities where available.
- ChatGPT Custom Instructions — recurring explicit preferences and settings.
- Memory FAQ — saved memories, chat history and Temporary Chat.
- Projects in ChatGPT — project instructions, files, chats and context.
- Data Controls FAQ — managing how conversations are used.
Lesson Summary
Seven ideas to remember
- Match every preference to the smallest useful scope.
- Use the current prompt for one-off instructions.
- Use Personality and Characteristics for communication style, not capability.
- Use Custom Instructions for stable guidance across chats.
- Use Memory deliberately and review saved details regularly.
- Keep project-specific rules, chats and files inside their Project.
- Use Temporary Chat for a clean conversation and always minimise sensitive information.