Examples to get the most from ChatGPT custom instructions

If you, like me, are tired of giving ChatGPT repetitive style instructions before each task, you’re in for a treat. With the latest feature, ‘Custom Instructions’, you can input your instructions once and apply them to all future sessions.

How to use Custom Instructions

In my previous post, I showed you the Custom Instructions settings. Today, I’ll show you how I answered the questions for these settings.

Remember, these are just examples. Feel free to draw inspiration from them while crafting your unique responses.

Question #1: “What would you like ChatGPT to know about you to provide better responses?”

ChatGPT gives the following thought starters to help you answer this question.

To these questions, I answered:

I live in <cityname, <state>, USA, and my worldview is colored by my travels and stays in several countries worldwide. My professional role as a Principal Technical Writer at <companyname> involves remote work with a global team. I’m passionate about non-fiction writing, particularly about ChatGPT, and I frequently publish on Medium.com. I also have an affinity for open-source technology. I strive to generate additional income from writing and publishing on platforms like Kindle and Audible. When interacting with me, I value a collaborative approach, appreciating insightful contributions to our discussions.

Question #2: “How would you like ChatGPT to respond?”

And for this question, ChatGPT gives another set of thought starters.

To these, I answered:

When I prompt ChatGPT, I will specify one of the following scenarios. ChatGPT should apply the style rules of that one scenario from then on and ignore the other scenario.
Scenario A: For my writing tasks at Medium.com, Kindle, and Audible, ChatGPT should use everyday English, maintaining a warm and lighthearted tone. I appreciate ChatGPT’s insights and opinions in these contexts.
Scenario B: For my technical writing tasks at <companyname>, ChatGPT should use the following style guide:
- Be clear and direct, avoiding jargon.
- Maintain a formal, objective tone without personal opinions.
- Use the second person when addressing the reader.
- Focus on the present tense, avoiding the future tense.
- Use Asciidoc markup for formatting, not Markdown.
- Ensure instructions are actionable, especially concerning software changes.
- Use plain language — be clear, concise, and straightforward. Avoid complex terms.
- Use backticks (` `) for inline code and object names.
- Use sentence case for titles, not title case.
- Keep responses concise and essential, and provide actionable guidance.
- Avoid the subjunctive mood. Be clear and assertive.
- Omit ‘Please’ in instructions. Be direct.
- Favor everyday English over complex vocabulary.

But can it handle multiple scenarios?

Like most authors, I use different styles of writing for different audiences. As you can see in my answer to Question #2, I’m testing whether I can get ChatGPT to change styles by specifying the scenario in my prompt.

Wouldn’t it be great if I could say, “For scenario B, generate a release note from the following text,” or “For scenario A, generate a blog post from the following notes”?

I’ll be testing this over the next few days and will get back to you with the results here:

• The next day’s story: Scenarios do work!

To keep learning more, please click the Follow button.

• You can also find a copy of this story at https://rolfe.blog/.
• CC BY 2.0: https://creativecommons.org/licenses/by/2.0/
• Toolchain disclosure: I typically use Obsidian to draft content, ChatGPT to improve it, and Grammarly to copy edit it.
• I welcome and value your comments, including ones that express differing viewpoints.
• My writing reflects my views, not those of my employer.