How to Assign ChatGPT Specific Personality Traits

You can shape ChatGPT’s tone, communication style, and preferred way of responding by describing the traits you want in your instructions. Set them in Custom Instructions for chats in general, or add them to a particular conversation when you want a temporary style; no special tools are required.

What You’ll Need

Access to ChatGPT and, if you want preferences to carry across chats, access to its settings.
A short list of traits and examples of the behavior you expect. For instance, “warm and concise” is a starting point; “use a friendly tone, keep answers under five sentences when possible, and explain technical terms” is more actionable.

Step-by-Step

Decide where the personality should apply. For a one-off task or role-play, give the instructions in that conversation. To use preferences across conversations, open ChatGPT’s settings and look for Personalization and Custom Instructions. The wording and location of controls can differ by app or change over time. If you cannot find them, check the app’s settings or help resources.
Choose a few traits. Pick two to four qualities that work together, such as patient, direct, curious, and encouraging. Avoid a long list of adjectives that overlap or conflict. “Blunt but gentle” can be interpreted in different ways, so explain what that means to you.
Translate each trait into observable behavior. Specify tone, length, structure, and how the assistant should handle uncertainty. For example: “Be patient: explain unfamiliar ideas in plain language. Be concise: lead with the answer and use short paragraphs. If information is uncertain, say so rather than guessing.” These instructions give ChatGPT practical choices to follow.
Set boundaries and priorities. Say what the style should not do and what matters more than style. For example, ask it not to use excessive enthusiasm, forced jokes, or unnecessary praise. Make clear that accuracy, clarity, and appropriate safety limits matter more than maintaining a character voice.
Write the instructions in a usable format. Put the most important directions first. You can use a compact paragraph or bullets, such as:
“Use a calm, friendly, practical tone. Start with the direct answer, then give numbered steps when explaining a process. Keep explanations brief unless I ask for detail. Define specialist terms. Ask a clarifying question if a missing detail would change the answer. Be transparent about uncertainty; do not invent facts.”
Save the instructions or send them in the chat. If using Custom Instructions, enter the text in the relevant field and save or apply it if the interface asks. If you are setting a temporary personality, send it as part of your prompt before the task. You can also include a reminder at the end of a long prompt if the style is especially important.
Test it with a normal request. Ask a question similar to the ones you expect to ask regularly. Check whether the response matches the tone, detail, and format you described. A single answer may vary, so test more than one kind of request before deciding that the instructions need a change.
Adjust one thing at a time. If the answer is too chatty, specify a shorter limit or say to skip introductory remarks. If it sounds cold, request considerate wording without excessive compliments. Change a specific instruction, save it, and try again so you can see what made a difference.

Common Mistakes

Using adjectives without examples. “Be smart, fun, and helpful” does not explain what response to produce. Add concrete guidance: “Use one light joke at most, and only when it fits; explain your recommendation in three bullet points.”
Giving contradictory directions. “Be extremely thorough” and “always answer in one sentence” conflict. Choose a priority or state when each applies, such as “Use a short answer by default; give more detail for complex topics.”
Making the instructions too long. A large persona description can bury the key preferences or include rules that are hard to apply consistently. Keep the most useful directions and remove repetition.
Expecting a perfect imitation or fixed character. Personality instructions guide response style; they do not guarantee identical wording every time, create human feelings, or make the assistant a real professional or person.
Assuming every setting works in every context. Customization controls and their availability can vary. Also, a request in the current chat may specify a different tone, so be explicit about which instruction should take priority.
Using a persona to replace accuracy. A confident or playful voice should not require fabricated certainty. Ask for honest uncertainty and reliable answers, and verify important information independently.

Tips

Describe what you want the response to do, rather than only naming a personality. “Acknowledge the question, then give a practical next step” is clearer than “supportive.”
Use a sample answer to demonstrate a style you like. Explain what to copy—such as short sentences or clear headings—rather than asking it to reproduce a person exactly.
Separate default preferences from task-specific requests. A useful default might be concise and plainspoken; a particular prompt can ask for a detailed tutorial or a formal email.
Use conditional directions to avoid one-size-fits-all replies: “Use bullets for several steps; use a paragraph for a simple definition.”
Review personalization settings occasionally. If your preferences change, update or remove old directions so they do not compete with new ones.
Keep private or sensitive details out of customization text unless they are genuinely needed. Style preferences usually do not require personal identifiers or confidential information.

Troubleshooting

The tone is still too formal or too casual: Replace broad labels with specific wording and a small example. For instance, say “Use everyday language and contractions; avoid corporate phrasing” rather than only “be casual.”

ChatGPT ignores a preference: Check that you saved the setting and that it appears in the correct personalization field. For an important request, repeat the key preference in the prompt. Simplify any conflicting instructions and put the highest-priority one first.

The answers are too long: Give a default length or structure, such as “Answer in three short paragraphs or fewer unless I ask for a full explanation.” Avoid demanding a strict word count for every task if some answers need more context.

The style changes between topics: Add examples for the situations where it matters, but allow the tone to adapt: “Stay friendly and clear; use a more formal style for professional messages.” Test with different task types and refine the instruction that is not carrying over.

You cannot find Custom Instructions: Look in the current app or website settings for personalization controls. If the option is not available in your interface, use a brief style instruction at the start of each conversation instead.

Frequently Asked Questions

Can I give ChatGPT a personality in one conversation only?
Yes. Describe the tone and response habits in your prompt. Those directions apply to the conversation, but may not carry over to new chats.

How do I make the personality apply to future chats?
Use Custom Instructions or the available personalization settings, then save your preferences. Controls can vary across interfaces.

Can I combine several traits?
Yes. A few compatible traits work best when you explain their practical meaning, such as being friendly while keeping answers brief and direct.

Will ChatGPT follow my instructions every time?
Not perfectly. Clear, relevant instructions improve consistency, but individual responses can vary and other requirements may take priority.

Can I ask it to sound like a specific person?
A safer, clearer approach is to describe general qualities—such as witty, concise, or formal—and provide a short example of the style you want, rather than relying on a person’s name.

Do personality instructions change what ChatGPT knows?
No. They guide how responses are presented; they do not add knowledge or guarantee that an answer is correct.

General steps for current mainstream versions; menus can differ by device and version. See the disclaimer or report an error.

More in this bed

The whole Tech How-To bed →