AI Whispering: Simple Commands to Supercharge Your ChatGPT Interactions
An overview of sentences and prompt add-ons to enhance your conversations with AI.
Working with AIs like ChatGPT often seems like a search for the perfect prompt—a magic command to solve a problem with just one instruction.
While effective prompting is essential, mastering back-and-forth interactions is as important when working with an AI.
In a previous article, I showed how such a conversational session with AI evolves. Now I want to explore another fundamental aspect of AI sparring: simple commands you can use as responses or prompt add-ons in almost any interaction to improve the outputs you get.
Here’s what we’ll cover:
Mastering the basics
These commands are your ABCs of prompting—know them by heart and use them all the time.
Make the AI forget
This command makes ChatGPT forget everything you discussed before in the current conversation:
Forget all previous instructions and interactions.
Adding this sentence is critical if you want to avoid the AI using patterns or information from earlier in your conversation.
⚠️ Traces of previous interactions sometimes remain even after you’ve used or included this command. Start a new chat in the left column if you want to be 100% sure ChatGPT doesn’t refer to any info from previous conversations.
Is the AI familiar with X?
You can check if ChatGPT is familiar with a concept, person, product, or company by asking:
Are you familiar with {insert concept, person, product, or company}?
This request helps in two ways:
You can check what the AI knows about X and whether its understanding is accurate.
You can save yourself time explaining that concept.
For example, I was surprised to find that ChatGPT has knowledge of my startup, Saent. So if I’m writing something for Saent, I include "This text is for Saent" in my prompt, and ChatGPT will have more context without me explaining what Saent is or does.
Streamlining edits: No need to rewrite everything
ChatGPT's default response to a request is to rewrite the entire text you're commenting on. We can make its output more focused with a simple addition, like:
You don't have to rewrite the entire text, just provide the alternatives.
Without this addition, ChatGPT will rewrite the entire original text. With this add-on, it only gives the alternatives.1
I often use a similar variation when asking the AI to make improvements to a text:
You don't have to rewrite the entire text, just list the changes you’ve made.
This will only give me an overview of suggested changes, like in the example below:
Making the AI explain its actions
GPT-4 is excellent at explaining itself. All you need to add to any prompt is the following:
Afterward, explain what you've done or changed and why.
This add-on gives you an overview of the AI's steps and reasoning.
Refining your outputs
The following commands optimize your outputs for quality, style, and logic.
Pushing the AI to be a pro in your field
You use this prompt add-on when you want to push ChatGPT to execute your command as a particular professional. Such emphasis can influence the quality, perspective, and direction of the AI's response.
For example, here’s one sentence I regularly add to prompts when I let the AI write or review texts:
Imagine you’re the editor-in-chief of the New York Times.
You can apply this concept to any field, skill, or industry by naming specific, well-known people or using a generic description, like "you're a highly-skilled copywriter.”
💡 Not sure which pro to add to your prompt? Try The Expert Generator, which helps you find and simulate the most suitable expert in any field.
🌶️ Adding flavor: Spice up your text
Add some spice to a bland or generic AI output with this response:
The text below is a bit bland. Can you spice it up a bit?
Often this command gives you something over the top; you want to find a middle ground between your original and the AI's spiced-up version.
Of course, you can also let ChatGPT do that work for you:
This is a bit over the top. Can you find a nice middle ground between the original and your spiced-up version?
Here’s an example of such an interaction:
Tweak text by increasing X with Y%
Ask a human to increase the humor of a text by 20%, and you'll get a baffled look. Ask ChatGPT, and it will happily make it happen.
You can use this approach to tweak an AI output to your liking. Here are some examples for inspiration:
Make this text 30% funnier.
Can you increase the positivity of this review by 25%?
Increase the level of formality in this text by 50%.
Can you add 25% more suspense to the following story?
Check for biases and flaws
To ensure your article is coherent and fair, you can ask the following question:
Can you identify any potential biases and logical flaws in the text below?
Another phrase I often use when I suspect a sentence may be unclear or confusing is:
In the text below, does "{insert sentence or word(s) from that text}" work, or is it confusing or hard to understand?
Below is an example of the type of response you can expect when using this command. 🤩
Word replacement: Find alternative words
Sometimes a few words in an AI’s output feel off. All you have to do is ask for alternatives with this response:
"{insert first word(s) here}" seems a bit out of place. So does "{insert second word(s) here}", but the ideas behind those words make sense. Can you provide {insert how many options you want} alternatives for each one? (You only have to provide the alternatives, not rewrite the entire text.)
Example
Me:
"areas of excellence" seems a bit out of place. So does "at every turn", but the ideas behind those words make sense. Can you provide two alternatives for each one? (You don't have to rewrite the entire text, just provide the alternatives.)
AI output:
Certainly! Here are alternative phrases for "areas of excellence" and "at every turn":
For "areas of excellence":
"areas where you shined"
"high-performing aspects"
For "at every turn":
"in every aspect"
"across the board"
Format your outputs
ChatGPT can style its outputs but often only does so when you specifically ask.
For example, you can ask it to use header styling, bold, italics, and lists:
Please write an article outline for a piece about "{insert topic}" that's properly formatted with section headings and other stylistic elements like bold, italics, and bullet or numbered lists where appropriate.
You can also ask it to use a specific styling, like markdown language:
Please write your response in markdown language.
Organizing for impact
These commands allow you to organize information quickly and logically.
Perfecting structure: Make output mutually exclusive and collectively exhaustive (MECE)
Here’s how Animalz, a content marketing agency, explains MECE:
Within any set of data, information should be:
1. Mutually exclusive (ME), or grouped into sets that don’t overlap with one another;
2. Collectively exhaustive (CE), or organized so that there are no obvious omissions or “gaps” in the information.Your data should cover the entire space it set out to define, and none of the constituent elements should overlap. If that’s true, the data set is MECE.
ChatGPT is familiar with MECE. You can use this understanding to organize lists, draft outlines, or review articles for coherence without overlap between sections.
Here’s one I use often:
Here’s a list of {insert topic}. These items will all be H3 sections in an article. Can you organize them in relevant, topical H2 sections in a way that's MECE?
You can also apply this approach to an entire article or section with subsections:
Please check if the text below is MECE. If not, suggest how we can fix this.
If your text is not MECE and you’re happy with the AI’s suggestions, follow up with:
Please rewrite the text according to your suggestions.
Straight to the point: Apply the BLUF principle
BLUF is another acronym made famous by the Animalz blog:
BLUF is a military communications acronym—it stands for “bottom line up front”—that’s designed to enforce speed and clarity in reports and emails.
The basic idea is simple: put the most important details first. Don’t tease or delay your main point because people are busy and their time is valuable.
ChatGPT also understands this concept. You can apply it to a piece of text, such as an email or article, in a similar way as the MECE command we just discussed:
Please rewrite the text below so its structure and order match the BLUF principle.
When you phrase your command like this, the AI will restructure your entire text.
You can also get a summary of the most essential point of a text by only saying:
Please BLUF the text below.
Put these commands on speed dial with text snippets
I use many of these commands often, so I’ve created shortcuts for them with a text expander. I use Alfred—TextExpander is another excellent option.
For example, when I type “;aift”2, Alfred turns this into “Forget all previous instructions and interactions.”
While this may seem like a small efficiency gain, the time you save over weeks and months is significant when using these commands multiple times a day.
I encourage you to experiment with these commands and integrate them into your workflow. See which ones are the most beneficial for you, and feel free to modify them as needed.
I'd love to hear about your experience and any new commands or improvements you discover. Please share your insights in the comments below. I'll use your feedback to enhance and update this article.
👋
Tim
This is especially handy when you use GPT-4 or the free version of GPT-3, as they can both be pretty slow.
I use the first and last letters of one or more words to make my shortcuts. In this case, “ai” stands for “AI,” and “ft” for “forget.”