Make ChatGPT feel like yours

Use four separate controls for standing rules, evolving context, broad tone and fine-grained response style.

InstructionsMemoryPersonality

Give each control one job

📌

Custom Instructions

State your role, recurring constraints, preferred structure and rules you want applied deliberately across normal chats.

🧠

Memory

Let useful context develop over time. Review, correct or delete remembered information when your situation changes.

🎭

Personality

Choose the broad tone ChatGPT uses when speaking with you. Task-specific writing instructions can override it.

🎚️

Characteristics

Fine-tune warmth, enthusiasm, brevity, headers and lists without changing the underlying capability.

A practical setup

Write standing rulesKeep them explicit, short and current.

Enable memory deliberatelyUse it when repeated context is helpful.

Choose a base stylePick the personality that feels natural.

Review periodicallyRemove stale facts and conflicting guidance.

🔐

Stay in control

Use Temporary Chat when you do not want a conversation to use or update memory, and avoid storing sensitive information simply for convenience.

How the controls interact in a real conversation

The current prompt has the most immediate influence, while standing settings provide context and style around it.

💬

The current request comes first

A specific instruction in the current chat can override your normal style for that task. For example, a request for a formal board memo can outweigh a generally casual personality setting.

🧠

Memory changes over time

Memory can maintain a continually updated synthesis of useful context rather than only a static list of facts. Review the Memory Summary because old roles, preferences or projects can become inaccurate.

🧹

Removing information may take several steps

Turning memory off does not delete earlier chats. To fully remove a detail, you may need to remove every place where it exists, including chats, files, memory and connected apps.