Enrico SchaeferFouning Attorney, Litigator and Ousdie Counsel - IP, Tech, TRAVERSE LEGAL, PLC · Founding Partner @ TRAVERSE LEGAL, PLC | AI, Blockchain, Social Media, Internet

3 Tips For Better Prompting in ChatGPT, Claude, Gemini & Perplexity

26 Nov 2025 · 3 min

Watch on YouTube

3:01

In this instructional video, Enrico Schaefer presents a three‑step prompt engineering sequence designed to improve the quality of outputs from large language models such as ChatGPT, Gemini, Claude, and Perplexity. He begins by emphasizing the importance of clearly defining the user’s intent and the desired “end zone” before any model interaction. The first step is to state the goal – in his example, drafting a notice of representation for a client. The second step involves pausing the model, uploading background information, and instructing the model to ask clarifying questions before any drafting begins. This ensures the model gathers all necessary context and addresses any ambiguities. The third step is the actual drafting phase, where the user provides specific guardrails (e.g., preserve evidence, outline general causes of action, request employee files under the Bullard‑Ploiecki Right‑to‑Know Act) and the model generates the notice. Schaefer wraps up with a friendly Thanksgiving sign‑off and a call to share, comment, and like the video.

Chapters

Prompt engineering is not dead. In fact, it is alive and well. Here is a prompt sequence that I think is going to really power up your output, the quality of your output, using ChatGPT, Gemini, Claude, these other tools, Perplexity.ai. The first thing that you want to do is really focus in on intent. What is your intent? What is the end zone that you're trying to get into? Where do you want to end up? And that context is going to be really important for the AI tool. So there's going to be 3 things that I want to teach you today. One is be clear about the end zone. Here's what we're trying to do. We're going to do a notice of intent or a notice of representation for a client. I'm going to give you the background information for that client. We're going to draft this notice of representation, but— Before we begin, I want to give you some background information. So step one is what is your intent? I'm going to do a notice of representation. Step two is to now dial back the AI by saying I don't need you to do anything yet. I'm just going to give you some background information and contacts so you could upload some of that information. You are then going to tell it. Specifically to ask you questions. Before we begin any drafting, I want you to ask me any questions you might have that'll help us prepare for the drafting exercise. Right? So you've given it the end zone. You've told it to hold off on any work. You've asked it to ask you some questions based on some some uploaded information. And great, now it's going to ask you some information for more information based on the upload. It's going to hold off on any drafting. After you've answered the questions that it has come up with based on your, based on your uploaded information and your end zone, your intent of where you want to end up, now you can say let's start drafting in Canvas. Here are the things that I want to convey. And look at it, a notice of representation could be a lot of different things. So you have to put guardrails on that notice. I don't want you to get very specific about the causes of action. I do want you to remind them about our prior communication to preserve evidence. I do want you to convey generally the causes of action that we're investigating. I do want you to demand the employee file under the Bullard-Ploiecki Right-to-Know Act. These are the types of things you need to include in your prompt. Until next time, my name is Enrico. Have a really happy Thanksgiving. We'll see you. Share, comment, like. We're on