NotebookLM Sucks, Until You Use it like this!

The first time I used NotebookLM, I uploaded a few PDFs, asked it a question, and got back a generic block of text that literally any AI chatbot could have given me. And just because of that, I almost trashed the whole tool off. But after 3 months and over 80 notebooks later, it's now the first thing I open when I start any research project. And the only thing that changed was how I set it up. The reason NotebookLM feels underwhelming out of the box is that it treats every notebook the same way. It doesn't know what you're trying to accomplish, who you are, or what kind of answers you need, so it defaults to broad and safe responses that technically answer your question, but never feels specific enough to actually use. So I'm going to create a new notebook and walk you through setting it up properly from scratch. And the very first thing worth mentioning is that the name of your notebook actually matters. Don't call it research or stuff or leave it as untitled. NotebookLM uses the name to help contextualize your sessions, and a clear name keeps your workspace organized as you build more notebooks over time. I'm putting together a freelance pricing strategy for a client, so I'll name this notebook Freelance Pricing Strategy Q2, so every time I open it, I know exactly what it's for and the AI has context from the start. Now before I start uploading sources or asking any questions, there's a setting that makes the single biggest difference in the quality of your results, and it sits in the top right corner of the chat panel. There's a small icon that opens configure chat, and inside you'll find a section that says define your conversational goal, style, or role. By default, this is set to default, which is exactly why everything comes out generic. I'll click custom, and a text field opens where I can tell NotebookLM exactly what role it should play when it responds. So for this pricing strategy notebook, I'll paste in a role. Now I'll ask a question with this custom role active, and then I'll switch it back to default and ask the exact same question on the same sources so you can see what changes. The default response gives me a clean, accurate summary of the sources. It reads clearly and cites everything properly, but it feels quite generic. It comes across more like a standard textbook style explanation than something tailored or insightful, so it doesn't really go beyond what's already in the material. However, the custom response isn't just summarizing. It's interpreting the sources through a specific lens, pulling in more targeted points, and organizing them in a way that feels more like advice from an expert. And while you're in that settings panel, there's another option right below it that's easy to miss, which is the response length toggle. For notebooks where I'm doing deep research or building deliverables, I set this to longer because I want the AI to go into detail and pull from as many sources as possible. And for notebooks where I just need quick answers or a daily summary, shorter keeps things tight and cuts out the extra stuff. It's a small toggle, but it changes the feel of every response in that notebook. Those two settings take about 30 seconds to configure, and they stay active for every conversation in that notebook until you change them. By the way, a huge thank you to Higgsfield for sponsoring this video. Higgsfield is the best platform for AI video creation right now because it puts every top model under one roof. Instead of paying for five different subscriptions just to test things out, Higgsfield gives you access to all the best generators in a single dashboard where you can compare results side by side. Whether you are making content, ads, or just experimenting with what AI video can do, Higgsfield makes the whole process faster because everything lives in one place, so you can seamlessly switch between state-of-the-art models like Sora 2, Kling, VEO 3.1, and everything else. I have left a link in the description below if you want to check it out. Now that the notebook is configured, the next thing that decides the quality of your results.