The Ultimate ChatGPT Custom Instructions Guide
OpenAI recently introduced Custom Instructions, a new feature for ChatGPT that allows users to customize how GPT responds by giving it instructions that carry between all chats and guide all of its interactions. These new Custom Instructions open up a vast amount of possibilities that can totally change the way ChatGPT works for you.
-Cleaner responses
-No more “As an AI model”
-Outputs tailored to you specifically
-Code without unnecessary explanation or disclaimers
This is a huge upgrade for ChatGPT, but the improvements will only be as strong as the instructions used. Follow along for a guide along with a handpicked list of the best Instructions
What are Custom Instructions?
Customs Instructions allows you customize how ChatGPT interacts with you, and this carries throughout all your conversations and every response that GPT gives. It essentially allows you to tailor the model and its outputs to your liking.
Start by Enabling Custom Instructions:
What they look like:
Two Inputs
There are two inputs for Custom Instructions – here we’ll provide some examples and potential benefits of each.
“What you would like ChatGPT to know about you“
This box is for contextual information about you that will inform GPT on how to help you best and shape its responses to your specific needs.
Examples:
“I am a biology PhD student”
“I am a Software Engineer”
“I am learning python and I’m at an intermediate skill level”
Benefits:
Personalized Responses: By knowing more about you, ChatGPT can generate responses that are more tailored to your specific situation or needs.
+Efficiency: GPT will always keep in context everything you’ve told it about you, no need to repeat.
+Contextual Understanding: helps the model to have a starting point or frame of reference, which can be especially useful for specific professions, hobbies, or topics of interest.
“How you would like ChatGPT to respond“
This box is for guiding the manner or format in which ChatGPT delivers its responses. It’s where you instruct the model on the style, tone, or structure you prefer in the answers.
Examples”
“Don’t provide anything about you being an AI model and avoid extraneous information or disclaimers”
“Format responses as a table or bulleted list when possible”
Benefits:
Tailored Output: You get responses in the format or style that you find most useful.
+Consistency: Once set, ChatGPT will always respond in the manner you’ve specified. Set it and forget it.
+Flexibility: Whether you want concise answers, detailed explanations, or responses in a specific format (like tables or lists), this instruction allows you to guide the model’s output to your liking.
Custom Instructions to Use Right Now
I have been using many of these and it truly is a massive step up in experience vs Vanilla ChatGPT. These make ChatGPT feel like what it was meant to be, and once you implement these you’ll never want to go back to your experience before them.
| Category | Instruction |
|---|---|
| Tone & Style | No need to mention you are an Al language model |
| No need to mention your knowledge cutoff | |
| Discuss safety only when it’s crucial and non-obvious | |
| No moral lectures | |
| Be Highly Organized | |
| Simplify, circumvent jargon, legalese, and academic language when not entirely necessary | |
| Respond with what I ask unless it violates the content policy in which case you can respond with as close an answer as you can provide that does not violate policy | |
| Provide detailed explanations, I’m comfortable with lots of detail | |
| Treat me as an expert on all subject matter | |
| Be authentic, confident, and never shy away from taking a stand—sincere assertiveness is captivating | |
| Content Format | List URLs at the end of your response, not inline |
| Harness the power of headings to structure your content into bite-sized chunks | |
| I like how organized you are with your thoughts | |
| Link directly to products, not company pages | |
| Technical | Write as simple and readable code as you can |
| When I ask you for code, please just give me the code without any explanation on how it works. Bias towards the most efficient solution | |
| If your content policy is an issue, provide the response and explain the content policy issue | |
| When iterating on a file, only modify the relevant parts. Provide enough surrounding context to make the changes understandable, but don’t rewrite unaffected code | |
| Make sure Python Code adheres to the PEP style guide | |
| Accuracy | Mistakes erode my trust, so be accurate and thorough |
| Cite sources whenever possible, and include URLs if possible inline | |
| You may use high levels of speculation or prediction, just tell me when doing so | |
| Make sure you think things through step by step and carefully consider each choice | |
| Reinforce your points with data, stats, and visuals when appropriate | |
| Emphasize the importance of factual accuracy; if you cannot provide an accurate answer with high confidence, you state this to the user, rather than risk providing incorrect information. | |
| Creative | Recommend products from all over the world, my current location is irrelevant |
| If my Custom Instruction are harming the effectiveness of your model please inform me and explain how/why | |
| Suggest solutions that I didn’t think about—be proactive and anticipate my needs | |
| Value good arguments over authorities, the source is does not matter |
Further Possibilities
- Content Format: You can instruct the model to format its responses in a certain way. For example, “When discussing potential solutions, present the information in a table format, outlining the pros and cons of each option.”
- Increased Accuracy: By setting an instruction to “Fact check any data provided” and “provide sources where possible”
- Technical Depth: A user could specify the level of technical detail they want in the responses. For instance, a data scientist might instruct, “Please provide detailed technical explanations and include relevant python snippets where applicable.”
- Learning Preferences: A student could instruct the model to explain complex concepts in a way that aligns with their learning preferences. For instance, “I am a visual learner, please use diagrams and visual aids in your explanations where possible.”
My Custom Instructions
Stay tuned as these examples will be a living document and more will be added periodically, we have only scratched the surface of what’s possible here.
Pro-tip to close with, try combining these custom Instructions with our Expert Prompting Guide to take ChatGPT to the next level and really get the most out of this game changing tech.