AUTHOR’S NOTE: I initially wrote this crash course intending to work with ChatGPT. But since then, Google’s Bar, Microsoft’s Bing AI, and many other AI tools have entered the market, making GPT less attractive. The principles I’ve explained below, however, can be applied to all AI chatting/generation tools – it’s prompt engineering so the rules won’t change no matter how diverse or advanced AI gets. So go ahead & read through!

In the last two parts, we covered some basic starter prompts to begin our conversations, proceeding to dive into deep research by chatting with AI.

Moving forward, we’ll get into specific use cases & examples of how to continue refining the basic response that GPT generates for you.

This is part of my larger crash course on ChatGPT, which is divided into 7 short parts:

1. Basic Starter Prompt Examples – get a feel of how to talk to an AI
2. Intensive Research Audit – use AI to mine insights & get information
3. Idea Generation – Get thought starters to kickstart your project
4. Refining Draft – Ask for feedback to improve your own thoughts/work
5. Content Creation – Develop original content from references that you liked
6. Team-wise Use Cases – List of prompts that each department might find handy
7. Beyond ChatGPT – Curation of other AI tools that can get your work done

Let’s start with our first case study.

How to Get Specific & Drill Deeper 

Everyone says you need to get super clear with your prompts. But how exactly do that? And how to push ChatGPT to give you better answers?

Let’s take an example. You can see my entire chat saved here – mind you, this entire exercise took me only a few minutes … that’s hours of hard work reduced to 5% of the time it would’ve taken to complete it.

---

Me: I’m making Instagram videos for an ad agency. Our target audience is 30-40-year-old marketing professionals who are brand managers. Generate 15-20 highly specific & niche topics that can be covered in less than 60 seconds. Cover subjects like SEO, marketing automation, web dev, UI/UX and email marketing. The topics need to be catchy, engaging & have strong hooks i.e. they should pull people in using questions, mystery, intrigue, shocking stats, etc.

Lesson: Notice how I got specific with my purpose (videos), platform (Instagram) and position (agency). I talked about my TG. I was clear in my ask – I started small (see how I take the conversation ahead).

ChatGPT: {Generates Response}

Me: I like idea no. 4 – How to write SEO-optimized blog posts. Generate the entire post for that. Assume the tonality of Rory Sutherland, an expert marketer with a quirky personality who uses British deadpan humour & puns in his speeches. Avoid using technical jargon & ensure the topics are practically relevant for brand managers. The goal is to showcase our expertise in these areas & increase credibility by giving useful insights so that managers are interested to work with us.

Lesson: Now I go one level deeper into what I want. I clarify my end goal & also a very niche tonality I want it to adopt. I also tell it what NOT to do.

ChatGPT: {Generates Response}

Me: This is vague. Make it more relevant for Indian audiences. Give more specific suggestions & use bullet points. Use examples to support each point. Keep your text concise & use line breaks often. Make it readable & snackable on mobile screens. Avoid using emojis completely & keep the formatting clean.

Lesson: If you aren’t happy with the response, you can always push it in the right direction. Clarify what’s wrong & what you need it to do better.

ChatGPT: {Finally gives me what I wanted}

Me: I then use this as a base template to write the actual post!

Key Takeaways

• When you’re trying to come up with one idea, say for a blog post that needs to go live today itself, start by generating lots of ideas – this helps you pick & choose the best one.

• When asking ChatGPT for ideas, give it as much context as you can on your goals, target audience, the time required to consume the content and what style you want your ideas to be (e.g. concise, funny, quirky).

• Once ChatGPT gives you a bunch of suggestions (and you can always ask it to generate better ones if you’re not happy), pick an option and THEN ask it to expand on that topic.

• When asking AI to expand on something, also tell it what not to do (e.g. don’t use visual cues or prompts in the script, just give me the words that I need to speak) so that it will skip any unnecessary clutter that’s not useful to you.

• Also focus on the kind of formatting, tone, style, etc. that it needs to respond in (as per your personal comfort level). The beauty of ChatGPT is that you can not only command it to give you answers but tune those answers in a way that’s easiest to consume for your case. Teach it your tastes & preferences so that it can adapt to them!

Navigate this Track

This is part of my larger crash course on ChatGPT, which is divided into 7 short parts:

1. Basic Starter Prompt Examples – get a feel of how to talk to an AI
2. Intensive Research Audit – use AI to mine insights & get information
3. Idea Generation – Get thought starters to kickstart your project
4. Refining Draft – Ask for feedback to improve your own thoughts/work
5. Content Creation – Develop original content from references that you liked
6. Team-wise Use Cases – List of prompts that each department might find handy
7. Beyond ChatGPT – Curation of other AI tools that can get your work done