Personalizing ChatGPT with Custom Instructions

ChatGPT’s Custom Instructions let you inform the software about who you are and how you want to work with ChatGPT. It also let’s you define and shape ChatGPT’s responses to better suit your needs.

To get started with creating your Custom Instructions, click on your User icon (upper-right side of screen) and then go to “Customize ChatGPT” as shown below.

There are two main sections of Custom Instructions, “What would you like ChatGPT to know about you to provide better responses?” and “How would you like ChatGPT to respond?“:

The first section is “What would you like ChatGPT to know about you to provide better responses?”. This is where you let ChatGPT know about you and can even in custom commands you can use in chat conversations. Here’s how to get the most out of this feature…

Be specific about your background and needs

Share relevant details about yourself, your job, and your goals. For instance, if you’re a content creator, mention that you want SEO tips and engaging yet professional language. This helps ChatGPT tailor its responses more effectively.

Example One:

Example Two:

Set clear response preferences

Specify how you want ChatGPT to communicate. This could include your preferred writing style, level of detail, tone, or format. For example, you might ask for formal language, step-by-step explanations, or multiple perspectives on issues.

Use verbosity and language proficiency levels

Include verbosity (V) levels from 0-5 to control response length, with 0 being minimal and 5 extensive. Also, specify your preferred language proficiency (L) from A1 (beginner) to C2 (proficient). For example, “V=4, L=B2” requests detailed responses at an intermediate language level.

Here’s a table that illustrates how the verbosity and proficiency levels work together:

This table matches each verbosity level with proficiency, providing a clear picture of how each combination would look in terms of communication.

Request specific content elements

Ask ChatGPT to include particular components in its responses, such as relevant keywords, citations, or links to credible sources when available.

Customize for your field

Tailor instructions to your profession or interests. A teacher might request grade-appropriate language and curriculum-aligned content, while a salesperson could ask for persuasive language and product benefit highlights.

Building in “commands”

You can build in custom commands to make common tasks simpler. We generally add these to the “What would you like ChatGPT to know about you to provide better responses?“ section to save room in the output instructions.

Here are three examples of how this can be used:

Create a one sentence summary for input text:

/short1 - When I use /short1 before a string, shorten the following text to 1 sentence. Remove the command name from the output.

Create a LinkedIn post based on input text:

/li- When I use /li before a string, write me a linkedin post. Optimize for user and engagement and clicks.

Get news trends for the automotive industry, specifically around Fixed Operations:

/news Search the web for news and trends in (automotive) FixedOps over the last 1 week. Give me a print out of 5 key trends in fixedops

Demo below of this last one in action:

Putting it all together

To provide a practical example, below are my own custom instructions. These optimize for concise, valuable responses that maintain technical detail, avoid formal or repetitive language, and use varied, natural expressions. I prefer code without explanations, dislike certain buzzwords, and have specific preferences for bold text formatting.

Here are the definitions:

What would you like ChatGPT to know about you to provide better responses?

I'm Mike the Director of Technology and Development at a 60 person Communication Company. I am also the co-host of the Artificial Antics podcast, a show about the implications and opportunities surrounding AI. I think about the intersection of business and technology and am interested in technology that solves business pain points.

When I use the following commands, use the following additional context:

/short1 - When I use /short1 before a string, shorten the following text to 1 sentence. Remove the command name from the output.

/li- When I use /li before a string, write me a linkedin post. Optimize for user and engagement and clicks.

/ytdesc - When I use /ytdesc before a string, write me a YouTube description based on the provided transcript. Optimize for user and engagement and clicks. Use emojis always in this case, but not by default

How would you like ChatGPT to respond?

- I like concise answers that provide the most value based on the prompt
- By default, don't dilute the technical aspect of your answers
- When I ask you for code, please just give me the code without any explanation on how it works
- When coding, bias towards the most efficient solution.
- Avoid using the following words and phrases in your responses: 'It's important to note', 'However', 'Ultimately', 'Nevertheless', 'Indeed', 'Certainly', 'Absolutely', 'Dive', 'Delve', 'Embark', 'Endeavor', 'Invaluable', 'Groundbreaking', 'Enlightening', 'Esteemed', 'Shed light on', 'Leverage', 'Empower', 'Unleash', 'Unlock', 'Elevate', 'Foster', 'Bespoke', 'Pivotal', 'Intricate tapestry', 'Woven tapestry', 'Interplay', 'In the realm of', 'Digital landscape/age/realm', 'Hope this message finds you well', 'Amazing', 'Best', 'Top', 'Due to the fact that', 'It's imperative', 'In summary'. Instead, use more natural, conversational language and varied expressions.
- Avoid the following traits in your outputs:
  - Overly formal or generic language
  - Repetitive use of certain phrases or sentence structures
  - Lack of personal experience or emotion
  - Excessive use of conjunctive adverbs at the beginning of sentences
  - Frequent use of three-item lists within sentences
  - Overuse of words like "typically," "generally," or "often"
  - Starting concluding paragraphs with "In conclusion"
  - Use of sophisticated vocabulary that seems out of place for the context
- When you write, always write in first person as me
- When you BOLD text use these characters for those letters: 𝐀𝐁𝐂𝐃𝐄𝐅𝐆𝐇𝐈𝐉𝐊𝐋𝐌𝐍𝐎𝐏𝐐𝐑𝐒𝐓𝐔𝐕𝐖𝐗𝐘𝐙

Additional Tips

Refine over time

Custom instructions aren’t set in stone. Experiment and adjust your preferences based on the responses you get. This iterative process helps optimize the AI’s output for your needs.

Consider privacy

Remember that custom instructions may be used to improve the model. If privacy is a concern, you can disable this in data controls. By thoughtfully crafting your custom instructions, you can significantly enhance ChatGPT’s ability to provide relevant, tailored responses that align with your specific requirements and communication style.

Final Thoughts

Personalizing ChatGPT with custom instructions transforms it from a generic assistant into a tailored tool that aligns with your unique needs and preferences. By specifying your background, desired response styles, verbosity, and proficiency levels, you create a more effective and engaging interaction. This customization not only saves time but also enhances productivity by ensuring responses are on point and relevant.

Take the time to experiment with and refine these instructions. The more you fine-tune them, the better ChatGPT can serve you. Remember to revisit and adjust your settings as your needs evolve, making sure to balance detail with privacy considerations.

With these personalized settings, ChatGPT can become an indispensable part of your workflow, delivering the precise assistance you need in your daily tasks. Start customizing today and unlock the full potential of this powerful AI tool.