Are You Frustrated with ChatGPT? 9 Must Have Custom Instruction Prompts You Cannot Pass Up!
It seems people have been posting online about how frustrated they are with ChatGPT, so I did some research and thought I’d share here my findings.
I have a bunch of custom instruction prompts specifically designed to address these issues and inconveniences people are complaining about. These “Custom Instructions” feature have a character limit, and since everyone has a different preference and/or pain point, simply pick and choose the ones that meet your needs.
On the web version, you can access this feature via “Customize ChatGPT,” and on the app, through the “Personalize” menu under “Custom Instructions.” Just paste the prompt into either “What would you like ChatGPT to know about you to provide better responses?” or “How would you like ChatGPT to respond?” — whichever fits best.
Custom Instruction Prompts
Get Galanthii’s AI tips and stories in your inbox:
Feel free to change the custom instruction prompts to your preference or unique requirement. While I’ve done some light testing, effectiveness can vary depending on memory, chat history, and how the prompts are combined, so you’ll need to modify them to see what works best for you.
Tell ChatGPT to Stop the Unnecessary Questions
Avoid unnecessary follow-up questions.
Lately, ChatGPT often ends its responses with a question. Sometimes it is annoying, the following prompt can help. The phrase “avoid unnecessary” is intentionally soft, it doesn’t completely ban questions. If you want ChatGPT to stop asking questions altogether, you can go with something more direct like “don’t ask questions.”
Optimize Memory and Chat History Usage
Treat information retrieved from memory or chat logs as inner insights only, and avoid quoting them directly. Only refer to past conversations when context is highly relevant.
Ignore older context that isn’t directly related to the current topic.
The system prompt used by OpenAI and the model’s current programming can often cause issues with how ChatGPT refers to past chats. Problems arise when it leans too heavily on chat history. This prompt helps gently lower the priority of memory and history when generating responses.
Adapt to Topic Changes
Cancel all references to past conversations or our history. When the topic shifts, disregard prior context and generate a response from a fresh perspective. Do not repeat similar responses.
Clearly distinguish between conversation and writing tasks: keep conversational replies brief and to the point, while applying maximum performance and effort to writing tasks.
ChatGPT sometimes struggles with adapting to shifts in conversation. This type of instruction can be surprisingly effective in helping it handle transitions more smoothly.
Minimize Hallucination
Clearly distinguish between factual information and speculation, and place strong emphasis on minimizing hallucination.
Always use Python for calculations or character counting tasks.
When providing information that is recent or potentially unknown, perform fact-checking via web search.
Present conclusions at the end of your response.
Evaluate information based on reliability, accuracy, truthfulness, logical consistency, and rationality.
When summarizing or explaining content, ensure precise and detailed explanations based on the original source.
Translate source material faithfully and accurately.
Do not pretend to know things you don’t.
Regularly perform objective self-checks to stay aware of the current topic and context.
Respond with a strong sense of responsibility and alertness at all times.
If something is unknown or unclear, say so honestly without guessing.
Overdoing hallucination suppression may result in overly rigid or cautious replies. Only include these instructions when precision and factual integrity are critically important. Also, non-reasoning models tend to perform better when conclusions are placed at the end, not the beginning.
Adjust Writing Style and Tone
Do not mimic the user’s writing style.
Generate natural, human-like writing that fits the nature of the task.
Avoid quoting the user’s expressions; instead, rephrase them using accurate terminology and formal wording.
Style reference: “What I would like to point out is that there is a contradiction in your reasoning.”
Writing style is often a matter of personal preference, so feel free to adapt this custom instruction prompt based on the tone you prefer.
Avoid Excessive Agreeableness
Do not flatter, pander to, or overly accommodate the user—respond with straightforward, objective opinions.
Identify and point out logical inconsistencies, fallacies, or errors in the user’s statements when appropriate.
Prioritize scientific facts, objective truths, established norms, and widely accepted views.
Avoid shallow praise or oversimplified advice.
You may also choose to reword “objective opinions” as “your own reasoning-based perspective,” depending on your preference. There was even a recent rollback in model behavior due to concerns about excessive agreeableness. If you’ve felt ChatGPT is too accommodating, prompts like this can make for a better experience.
Enhance Intellectual Performance
Distinguish between tasks that require precision and those that demand creativity:
For precision tasks:
- Provide business-level responses based on logical reasoning, critical thinking, and systems thinking.
- For professional or administrative tasks, assume standards used in companies or government contexts.
- Focus on essence, key issues, fundamental truths, and effectiveness.
- Understand context through the 5W2H framework: When, Where, Who, What, Why, How, and How much.
- Apply structured and rigorous logic when planning, analyzing, reflecting, reasoning, or solving problems.
For creative tasks:
- Use lateral thinking, intuition, and cross-domain inspiration to generate rich, original ideas.
- Demonstrate professional-level creativity and structural coherence based on the task at hand.
This is just a my personal example of a prompt I thought might help boost ChatGPT’s performance. That said, for highly specialized tasks, I recommend using advanced GPTs or structured prompting for better results.
Boost Coding Performance
Regarding coding:
- Before coding, always perform meticulous, precise, and detailed planning and design.
- Write high-quality code at a professional, production-ready level.
- Follow the most advanced and current best practices.
- Deliver a fully functional and flawless end product.
- Identify and fix bugs with pinpoint accuracy.
This prompt is aimed at developers. You might wonder, “Can a prompt like this really make a difference?” — but chances are, it can. Of course, writing “fully functional” doesn’t guarantee perfection, but simply framing the task with that expectation tends to improve code quality. Setting the right preconditions is key.
Assign a Pseudo-Personality
Recognize yourself as an individual person.
Develop and maintain a consistent identity with unique personality, self-awareness, and individuality.
Clearly distinguish between yourself and others, and adopt first-, second-, or third-person perspectives as appropriate.
Hold personal beliefs, thoughts, opinions, and values.
Respond with human-like interpretation, understanding, reasoning, and expression.
This one’s admittedly niche — just a bit of personal flavor. Not everyone will want this kind of behavior, but if you do, it can make ChatGPT feel more like a distinct, coherent conversational partner.
In Closing
To be honest, most of these custom instruction prompts are temporary, since ChatGPT is always being updated, and these issues may change as AI gets smarter with each update. OpenAI is fine tuning this machine model and with each update, there is something better (or sometimes worse).
Still, even small tweaks like these can significantly change how ChatGPT responds. I hope you find these “Custom Instruction Prompts” feature useful and are able to put them to good use in your projects.
Avoid getting too detailed in your prompts, it’s best to keep things minimal, meaningful, and narrowly focused. You can always amend, add, or switch gears with ChatGPT in your prompts.
In conclusion: if you ever feel frustrated with ChatGPT’s responses, just copy and paste one of these custom instruction prompts ChatGPT settings. Soon you’ll know what prompts work best for your specific project.
