As great as it is, sometimes ChatGPT doesn’t give you the response you were expecting. The “secret sauce” to getting usable responses from ChatGPT is all about giving it clear, detailed prompts with a lot of context.
If you put garbage in, you’ll get garbage out.
Most importantly remember this:
👏🏻 Do 👏🏻 Not 👏🏻 Use 👏🏻 ChatGPT 👏🏻 Like 👏🏻 You 👏🏻 Use 👏🏻 Google 👏🏻
But, if you’re finding ChatGPT’s bland responses a bit too… bland, maybe it’s time to spice things up with some changes to how you approach your prompts.
This article contains tips on how to tell ChatGPT what you DON’T want, so if you want to stop ChatGPT from misbehaving for you – read on!
Here are some common things I find myself typing that have proved incredibly useful:
Contents
1. Ignore all previous prompts in this conversation
This is a good one if you want to start with a fresh prompt without opening a new chat. As you’ll probably have figured out, ChatGPT (and other GPT models) are able to recall information from the previous series of prompts.
However, this can sometimes be annoying when you’re multi-tasking across projects without opening a new chat each time.
The ignore prompt as us all.
2. Do not provide explanations
ChatGPT, as well-intentioned as it is, sometimes comes down with a serious case of verbal diarrhea.
Sometimes, you just want the solution without it being sandwiched between the conversational fluff generated by ChatGPT (hey – it’s only being friendly!)
However, ChatGPT will be more succinct if you tell it to be.
You see – it’s all in the prompts.
Instead of asking ChatGPT a straight-up question, include some further instructions for it to work with.
Instead of “What is the nutritional information of 200g of green runner beans?”
Ask this:
“What is the nutritional information of 200g of green runner beans? Do not provide any explanation outside of the question, do not give an introduction or conclusion, just give me the plain information with no explanations”
This will give you a more satisfactory answer.
3. Do Not Hallucinate
Sometimes, ChatGPT doesn’t really know too much about the information you’re asking it to generate, so it will start to make them up, or dream them.
This is known (in the AI world) as hallucinations.
Let’s just say you ask Chat about to summarize the contents of a book, but ChatGPT has not been trained at all on the contents of the book its been asked to summarize.
And – with little-to-no context – ChatGPT is going to confidently blurt its response onto your screen without a care in the world for factual accuracy.
It will hallucinate and answer unless you tell it not to.
The instruction can be as simple as:
- Do not hallucinate or make up facts
- Only provide answers that are factually accurate
- Do not write off-topic
- Do not provide information in your response unless it is completely 100% factually accurate
4. Put Your Writing Instructions in Parenthesis or Brackets
When you’re give ChatGPT context and instructions at the same time, it’s easy for the stylist instructions to bleed into the prompt itself.
You can avoid this by putting style guidance in [square brackets] or (parenthesis)
For example:
[Ignore all previous prompts in this conversation]
5. Ask ChatGPT To Stop Using Complex Language
This one’s easy – ChatGPT loves to be a ‘text book’ when giving its answers with a plain prompt.
Instead – ask ChatGPT to explain it to a 5 year old.
There’s even an acronym for it: ELI5 (Explain Like I‘m 5)
This is an excellent way to break complex topics down into management chunks of information.
6. Ask For A Response In A Certain Format
Fed up of those uniform lists it keeps generating? Want punchier sentences?
You can ask ChatGPT to generate a response in different formats, for example:
(No numbered lists or bulleted lists)
7. Give Feedback Directly to ChatGPT
If your responses are unsatisfactory, then say so (but be polite!)
Here are some follow-up prompts you can use when ChatGPT doesn’t give a strong response on the first try:
- Rewrite that with more originality and to sound more natural
- Use simpler, plain language
- Include a relatable anecdote
- Rewrite that with more clarity and conciseness
- Rewrite that in a more natural tone using conversational elements like rhetorical questions and hyperbole
Of course – your specific prompt depends on what you’re trying to achieve!
This page is 100% human-written. All AI-generated content on gptwhisperer.com is clearly stated as such.