Is NotebookLM Still the BEST Deep Research Tool?

I tested the most popular AID research tools to find the best one and one that no one even knows about actually performed better than those. I took the same task and I gave it to ChatGpt, Gemini, Notebook LM, Perplexity, and even Claw. And none of them outperform this secret tool that I'm going to show you in this video. The reason why is because every single platform does the same thing. They go wide and bring in as much data as possible and then they synthesize it and call it research. But none of them hardly evaluate the source quality and they don't come near to applying any type of research disciplines or methodologies. But when I score each of their research papers against a professional standard, their differences were brutal. Most of the time when you think about research, the number one underrated tool that comes to mind is Notebook LM. Notebook LM was supposed to be this AI research intelligence, but lately with all of the updates they've been rolling out, it's beginning to be more of a competitor with Canva and less of an AI research tool. You can create infographics, slide decks, podcasts, and all other types of visual and audible content. But when it comes to research and new research features, it seems that the platform is at a stall. And that led me to ask one critical question. Was I ever really doing research? Most of the time, what we do is we upload PDFs or we do a search, we get some sources, and we start asking questions. and we think we're doing research, but in reality, we're just doing selective reading because Notebook LM is simply retrieving and presenting the information that's contained in the sources. And so, I began to understand that unless you have a structured approach with the questions that you're asking inside of Notebook LM, you're not really doing research. And that's why I built my Notebook LM research assistant wizard. It's a bonus inside of the 1hour content strategy. You can get it right now on my website if you want access to it. But essentially, this research wizard would turn any notebook inside of notebook LM into a true research assistant that actually gave you research questions in a progressive manner that allowed you to extract the knowledge and insights that you wouldn't get just by asking questions and reading responses. The number one problem that each of these platforms is facing is a problem with source engineering. And the only platform that's even begun to tackle this is OpenAI. When you do deep research inside of chat GPT, you can actually prioritize some sites over the others. And that's a huge advantage over the other platforms when it comes to deep research because it allows you to control the quality of your sources to some degree. But when I ran this test, I did not use that feature so that I could test the base models of each deep research evenly. So when it came to the research question of how are solarreneurs using AI to replace their first hire, many of these quote unquote sources were just blog posts promoting their own tools masquerading as research. And none of the platforms discriminated against them. They brought them in, treated them like they were verifiable truth, and presented them to me. If you're doing online research and looking for sources from the last few years since chat GPT was first released to the public, then you definitely need to make sure that you're auditing the research that you're getting from any of these tools because there are so many gurus teaching people how to flood the internet with AI content and it's mostly slop and it's coming into the deep research. If you want to use AI to do online research and you don't want to worry about whether the results you're getting back are professional grade or something that you can defend, then make sure you stick around to the end of this video because I'm going to show you a tool that outperformed them all. But I'm also going to give you some tips to make sure that you're not getting this slop into your research papers. Out of the seven deep research tools that I tested, Gemini came in last place with a score of 52%. Gemini built this entire economic thesis on vendor marketing blogs. 72% of the sources were low confidence. Gemini did everything it could to earn the lowest trust score in this experiment. And you guessed it, Notebook LM came in at number six with 59%. 53 sources, but the second lowest score of any platform. Notebook had a 72% low confidence sources in academic formatting. It looked rigorous, but it wasn't. There were just blog posts, nothing substantial, no backlinks, very weak. But it did have one novel framework that it came up with called the cognitive task suitability matrix, but it wasn't properly sourced. So, it was a good idea, but it still didn't have the necessary evidence to support that idea. So, that's something I'll be going back to visit and probably extracting later. Number five is a tool that I very rarely use, but I see a lot of people hyping up, and I just never got the hype. I don't do any videos about Perplexity AI. Perplexity had a 59% as well. zero high confident sources. Not a single one. None. But it was the only tool that didn't overstate what it knew. And so being honest about having no evidence is actually better than hallucinating and trying to push a bunch of marketing at me as research. Number four is not a platform. It's my own personal plug-in that I launched inside of Claude AI. But there are two things I want to let you know about this. Number one, it only came back with 12 sources. And number two, I understand that Claude is grading itself. So there

is a definite probability that it's being lenient with itself. But it was the first tool out of any of them that performed a self- auditing methodology. The plug-in didn't just gather information and then give me a bunch of statements and findings. It actually gave me a confidence score with each of those findings and it even gave me a gap report to let me know what it did not find out, what questions were still unasked, and what still needed to be researched in order to increase the confidence in the findings. And as great as I thought the plug-in was before this experiment, I was humbled because I found out that it was nowhere near as good as I thought it was. I used to think that Chat GPT didn't do enough deep research because it brought back a very small number of sources. But in reality, as I said before, they're one of the only platforms moving in the direction of quality over quantity. But even so, there were still a lot of instances where it overstated its findings. The second platform was actually Claw's own deep research that came in with the 79 and beat Chat GPT by only three points. Claude came back with a total of 65 sources. And Claude AI was the only deep research tool that caught a fake Mckenzie stat and debunked it. And of 65 sources, 11 of them have a high confidence of being true and good research papers. And this is the part of the video I think is more important than everything else. And that's more important than which platform you should use because I use my own librarian skills that outperform three of the major platforms in the deep research space. And what I did next was I simply asked Claw, why did my research plugin not perform as the well as the other platforms and Claw was able to compare my research data with their research data and find every little hole I had in mind, every little flaw. And so after I found it, I told Claw, "Listen, I want you to turn this into an implementation plan. " so that I can edit this plugin and make it better. And that's what Claw did. And then I told it to update the plugin. It updated the plugin. I saved it. I ran the research again. And this time the results were totally different. My Liberia 3. 0 plugin came back with a score of 82%. One of the reasons it scores so high is because it's going to make three passes on the internet to make certain that it has as much government, academic, highquality data as it possibly can before it even begins to look at a blog. 47% of the sources from this research assistant were tier one or tier 2. And when you compare that to Chat GPT's 20%, Gemini's 9%, Notebook LM 6%, and Perplexity 0%, you immediately see the difference. This is real research. And probably more important than anything else, this plugin doesn't perform a simple Google research. This plugin looks at the question that's being asked. And the next thing it does is it determines which of the 10 desk research methodologies it's going to use for that particular query. Instead of it giving you the questions to ask inside of Notebook LM, it asks the questions itself. It finds the data itself and it does the research for you. It truly is an automated research assistant. And so if I had to tell you one platform to use over any other for AI research in 2026, I would tell you to build your own with a claw plugin or a claw skill as that platform with co-work is especially set up to help you perform these type of tasks. And chatgpt if they ever install something similar to claw co-work will be able to help you do the same thing as well, but for now they don't have it. But Notebook LM is not the best research platform.