ChatGPT Personalization Guide for Beginners: Base Style, Tone, Characteristics, Custom Instructions, and Memory

Open ChatGPT's Personalization settings and you may see several controls: Base style and tone, Characteristics, Custom Instructions, and Memory. Their names sound related, so it is easy to wonder where to start or assume that every option needs detailed configuration.

What can ChatGPT Personalization do? The short answer is that it can adjust four different layers: the general feel of replies, finer presentation choices, explicit preferences you want followed, and recurring context carried across conversations. The beginner-friendly settings and examples in this guide start with one base style and one short custom instruction. Characteristics and Memory can wait until a recurring need becomes clear.

This guide uses OpenAI's official information available on August 5, 2026. It explains what each control does, how the controls interact, and how to build simple default, work, and learning setups without over-configuring your account.

Think of Personalization as four separate jobs

The controls in Personalization are not four versions of the same instruction. They serve different purposes:

• Base style and tone selects the general voice used for replies.
• Characteristics fine-tune qualities such as warmth, enthusiasm, formatting, and emoji use.
• Custom Instructions state your role, preferred output, and standing rules directly.
• Memory uses recurring context and preferences from conversations and other available sources.

OpenAI says these controls work together. The official Characteristics in ChatGPT guide explains that Characteristics retain the same capabilities while adjusting the vibe, and that they work alongside the selected personality, custom instructions, and saved memories.

The benefit of combining them is consistency: you may spend less time repeating that you want a concise answer or beginner-level explanation. The drawback is that overlapping or conflicting settings can make the result harder to predict. If you select Friendly, reduce warmth, and also instruct ChatGPT to always sound highly conversational, it becomes difficult to tell which control produced the final tone. Start with one change and test it before adding another.

Base style and tone changes how replies sound

Base style and tone chooses the normal communication style for ChatGPT. OpenAI's official ChatGPT Personality guide describes Default as clear and neutral and lists alternatives including Professional, Friendly, Candid, Quirky, Efficient, and Cynical.

Professional is a sensible starting point for polished workplace communication. Efficient favors concise, direct answers and is useful for checklists or troubleshooting. Friendly provides a warmer conversational style for reflection and planning. If none of those represents a clear improvement, leaving Default selected is completely reasonable.

A personality does not change what ChatGPT can do or alter its safety rules. It affects communication, not capability. It also does not control every deliverable. When you request code, an email, a résumé, or another specific artifact, ChatGPT may follow the requested format and immediate context more strongly than the selected personality.

On the web, OpenAI directs users to the profile icon, Personalization, and then Base style and tone. The corresponding option is under Personalization on iOS and Android. A personality change applies across chats, including existing ones. Labels and placement can vary with language, device, and app updates, so use the function's purpose rather than relying on one exact translated label.

Characteristics make smaller adjustments

Characteristics let you nudge individual aspects of a response in a more or less direction without replacing the entire personality. As of August 5, 2026, the official Characteristics guide documents Warm, Enthusiastic, Headers & Lists, and Emojis.

For example, you might like Efficient's directness but increase Warm if it feels too impersonal. You can reduce Headers & Lists when you prefer connected prose, or reduce Emojis for informational answers. These settings do not add knowledge or unlock abilities; they adjust how otherwise similar content is presented.

Characteristics are rolling out gradually and may not appear for every user yet. If the section is absent, it does not necessarily indicate a problem with your account or app. No official information is currently available to confirm a single date when every plan and region, including Japan, will receive it.

Custom Instructions store rules you would otherwise repeat

Custom Instructions are for information and rules that you explicitly want ChatGPT to consider. OpenAI Academy's Personalization guide gives examples such as your role and responsibilities, preferred tone, desired output format, and a guardrail to ask clarifying questions when requirements are unclear.

A beginner can start with only three lines:

I am a beginner in this subject. Briefly explain technical terms.
Lead with the answer and use numbered steps when there is a procedure.
If missing information would substantially change the answer, ask before guessing.

OpenAI's Custom Instructions FAQ says the feature is available on all plans across web, desktop, iOS, and Android, and changes apply immediately across chats. The most useful entries are stable preferences you genuinely repeat. Today's deadline, the word count for one document, or instructions for one uploaded file belong in the current chat instead.

For a longer template and a detailed explanation of what belongs in each section, see the existing guide to what to write in ChatGPT Custom Instructions.

Memory is for recurring context learned over time

Memory helps ChatGPT use relevant context from chats, files, and connected apps so that you do not need to repeat yourself as often. According to OpenAI's Memory FAQ, Memory can be enabled or disabled under Settings, Personalization, and Memory. The Memory summary lets you review and correct what ChatGPT remembers, although the summary may not display every factor used for personalization.

The distinction from Custom Instructions is useful. Put explicit standing directions, such as an output structure, in Custom Instructions. Memory is better suited to recurring context shared through conversations, such as your role, ongoing interests, or durable preferences. OpenAI Academy similarly recommends Memory for context that will matter again and advises against saving one-off details.

Convenience comes with a privacy trade-off. Sensitive information may appear in Memory if you share it. Do not enter passwords, private keys, confidential client data, or other secrets as personalization material. If you do not want a conversation to use or create memories, OpenAI says you can turn Memory off or use Temporary Chat. Temporary Chat neither uses existing memories nor creates new ones.

Deleting one visible memory may not remove the information from every source. OpenAI says that fully removing something can require deleting it from past and archived chats, files, the Memory summary, and connected apps where it appears. Before handling important information, ask ChatGPT what it remembers and inspect the Memory summary directly.

Three simple setups for beginners

You do not need every option enabled. Use one of these combinations as a starting point and only configure controls that are available in your account.

Default setup for understanding the effects

• Base style and tone: Default
• Characteristics: unchanged
• Custom Instructions: “Reply in English, lead with the answer, and briefly explain technical terms.”
• Memory: initially off; reconsider after you identify recurring context worth keeping

This setup makes cause and effect easier to see. Try three ordinary questions. Add another setting only when the same problem appears repeatedly.

Work setup for concise, reviewable output

• Base style and tone: Professional or Efficient
• Characteristics: more Headers & Lists and fewer Emojis
• Custom Instructions: “Organize replies as conclusion, evidence, and next action. Ask when a missing requirement changes the result.”
• Memory: limited to a durable role and general output preferences

Do not store customer names, contracts, credentials, or confidential project facts in Personalization. Follow your organization's data rules. Conditions that apply to only one engagement are easier to manage in that chat or its Project rather than as global preferences.

Learning setup for guided understanding

• Base style and tone: Friendly or Default
• Characteristics: more Warm and Headers & Lists when available
• Custom Instructions: “Explain at beginner level, give one example, then ask one question to check my understanding.”
• Memory: keep only the field being studied and durable learning goals

Avoid a standing instruction to always provide the final answer immediately or oversimplify every topic. A rule such as “show the reasoning approach first, then explain the answer when needed” preserves more of the learning process.

When a setting seems ineffective, remove conflicts one at a time

An unexpected tone does not necessarily mean the setting failed. OpenAI's Personality guide says a saved memory that conflicts with a personality can override or reduce its visible traits. Instructions in the current conversation can also adjust or obscure the selected personality.

Test the same short request in a new chat. Then check Custom Instructions for phrases pointing in the opposite direction. If the result still seems wrong, review tone preferences in Memory. If Characteristics are available, move them near their defaults and compare again. Changing one item at a time makes the source of the behavior easier to identify.

Personalization does not guarantee factual accuracy, and it does not eliminate the need for a clear prompt. Requirements that apply only now—checking current information, writing to a fixed length, or addressing a specific audience—should still be stated in the conversation.

Summary

ChatGPT Personalization separates four useful jobs. Base style and tone establishes the general voice. Characteristics adjust presentation and emotional tone. Custom Instructions store explicit standing rules. Memory carries forward recurring context from conversations and other available sources.

Beginners can leave Default selected and add one short Custom Instruction. If replies remain too long, try Efficient. If conversation feels too impersonal, try Friendly or increase Warm. If you keep repeating the same durable context, consider Custom Instructions or Memory. Keep secrets out of these settings and review Memory periodically.

Instead of configuring every switch today, remove one sentence you repeatedly type—the smallest change is often the clearest path to a genuinely personal ChatGPT setup.