What I learned about ChatGPT while creating a library of 300+ Prompts

If you’re a marketer, ChatGPT can be a scary word.

Even I was intimidated initially but I still couldn’t get all the hype.

I experimented with it by giving it a few basic commands (aka prompts), but the results were still not as good as the internet claimed.

That’s when I stumbled upon a few marketers on platforms like LinkedIn and Medium who seemed to have learned the art of writing prompts. I had my first learning after following their tips (and copying their prompts). Sharing the most important ones below -

ChatGPT is as good as the prompts you write

You can train ChatGPT to behave as a writer, proofreader, researcher, and editor if you know how to write clear prompts.

For example, I typed a simple prompt asking for blog ideas, and I got a few generic ideas that I could’ve come up with myself easily. (Although, it did save me 30 minutes of brainstorming ngl)

But when I used a prompt submitted by an expert on Prompt Hunt, I got an entire plan made, which otherwise would’ve taken me half a working day.

You can see the difference between the results. Over the past few months, I have learned clear, contextual, and specific prompts always lead to better results.

Don’t be afraid to demand more from ChatGPT — you want your results in a specific format, mention it; if you want it to omit certain words or references, state it in the prompt.

Honestly, it is all in the prompts.

Although there’s more to ChatGPT than writing prompts

At the risk of contradicting myself, I’d like to share that multiple settings on ChatGPT can help you get desired results, and they’ve nothing to do with writing prompts.

But what do these settings control exactly? Remember that ChatGPT is ultimately a machine. It converts your words into the format it understands and then acts upon them. This format is called tokens.

Here’s how you can control it -

1. Max tokens — It specifies the length of the output. A higher number of tokens results in longer responses, whereas a lower number of tokens may shorten the response (and even give incomplete answers, so keep it balanced).
2. Temperature — It helps you control the randomness of the output. The scale ranges from 0 to 1, and the lower range gives you formal output, whereas the higher range provides creatively diverse results. Here’s the breakdown —
   0.0 to 0.3 > Formal and non-creative writing
   0.3 to 0.7 > Balanced tone
   0.7 to 1.0 > Creative and diverse tone
3. Presence Penalty — Like its name, it penalizes the presence of tokens in the results when they’ve already been used in the input. Positive value discourages repetition of token, whereas negative values echo the input.
4. Frequency Penalty — This setting can direct ChatGPT to use rarer words instead of common vocabulary. Positive value promotes less common words and vice-versa.

I also found this detailed video that explains the “ hows” of all of this

ChatGPT does get better with a little help and, of course, a few Chrome extensions.

Browser Extension amplifies ChatGPT experience

The tech world didn’t waste time and has created hundreds of Chrome extensions to maximize ChatGPT’s output. My team and I have become a fan of a few of them, like —

1) Grapevine — Summarizes webpages

2) Merlin — Summarizes YouTube video, creates Google Sheet formulas, and suggests email replies.

3) TweetGPT — Changes the tone of any tweet based on your audience

Do try them out, and you’ll learn that ChatGPT DOES get better with extensions and integrations.

Yet, good prompts remain the key

What helps the most (in getting the best out of ChatGPT), however, is learning how to write the right prompts, and there’s no better way to do it than learning from experts.

Our AI prompt library helps you discover, submit, and rate marketing prompts from other experts.

It already has 250+ prompts from 50+ experts. You don’t have to skim through the dark abyss of the internet to learn about expert prompts (like I did) — we’ve already done the hard work for you, and people seem to be finding it useful already.

So, go check it out here if you haven’t.