Transcription
In this video, I'm going to show you a workflow that is going to completely transform the way you do research with AI. We are going to be using four different tools, Gemini, Cloud, Chat, GPT, and Notebook LM. And we are going to combine their outputs to share my secret AI research workflow with you.
We are going to be looking at a singular example, that of course you can then transpose and apply to your specific research challenge. Uh, in this case, we are going to be looking at an issue related to Google Analytics 4, and we are going to start with Claude. Uh, so in Claude, you just click on extended thinking so that it has more time to think and conduct better research. But then you go into research and tools and make sure uh to just switch this toggle to on. So we are in deep research mode right now. Then you of course need the right prompt, and you will uh ask it to conduct research about your specific issue. Actually, I haven't even explained this issue in such great detail, but of course, the more details you provide, the better. And here, the research has already been completed. And you can see that it has searched for 379 sources. And we have this huge "not set" traffic um report ready. But this is just the first step.
And now I'm going to show you how the real magic works. The next step is that you conduct deep research in another tool, which is Google Gemini. Uh, so again, here you go into the settings and then deep research mode, and then you give it the same uh type of prompt, or you can even provide a little bit more context. And then he, uh, Gemini will put together a research plan uh, which you can just um, see here. And I'm just going to say, "Start research." And then after a couple of minutes, uh, Google Gemini is going to conduct its own deep research. And you can see that it has actually searched um, also many different sources, and perhaps different sources than the ones that were used by Claude.
The third step here is that you go into Chat GPT, and in Chat GPT, you also uh, start the deep research. You just go into here, you click, click the plus sign, and then uh, you have uh, let's say, 25 deep researches uh, per month. Of course, this is on the Pro version or the Plus version. Um, most of this unfortunately needs a premium version. But then again, you use the same prompt. Uh, it will ask a few clarifying questions. You will just answer, and then again, Chat GPT is going to combine its own deep research. You have three different sources of this deep research, and they all have different types of information. They have searched the web in a different way because these AI models work in a completely diff, I mean, not completely, but they are configured differently. So the output of each deep research is going to be different.
So now, the next step in the secret AI research workflow is to be combining uh, these research findings into a singular source of truth. So this is exactly where Claude.ai 4.5 comes in, because it has this ability to just create huge amounts of output, especially compared to Chat GPT 5 or Gemini 2.5 Pro, which usually give you answers that are quite succinct. But here in Claude, of course, if you have the uh, right usage amount, you are able to create potentially a huge document. Uh, so this is where the secret sauce is. You basically first download these um, reports uh, from these uh, three different platforms. So you would just go uh, here into uh, copy, and then you download it here as markdown. In uh, Google Gemini, you just click on share and export, and then you can export it to Docs, and from Docs, you can just download it as markdown, which is a language that AI is very familiar with. And finally, in Chat GPT, you just click here on this arrow, and then you can just click on download, and also just download it as a Docs file or whatever, but then ultimately, you want to save it as uh, a markdown file. So, .md.
And then you come back to Claude, and as you can see here, I have these three separate reports that I have uploaded using this plus uh, sign over here, and then upload a file. So then they are already loaded here, and then uh, I have this kind of secret prompt that I can also copy into the description of this video, perhaps. Uh, "Please become a master report builder and researcher. Your sole task for this session is to collect all the findings from these three reports I have found about solving the 'not set' issue in Google Analytics 4. I have pulled these reports from different sources, and they all have unique insights, but they also have some repetitions. I would like you to use your full processing power to combine these three reports into a single unified document that will be very well organized. I want it to combine all insights from the three documents so I can have the largest amount of information to pull from when trying to solve the 'not set' issue. Gather all unique insights from all three reports. Organize everything into a logical sequence so that I can use this document as my ultimate guide that will allow me to reveal all possible causes of the 'not set' issue in Google Analytics 4. Leave no stone unturned and give it your best to create a perfect artifact for me. At the end, tell me what you've done."
And this prompt has uh, worked incredibly well. As you can see here, um, I don't have to jump between these three different reports that all have some of the same information, but also some unique information. Now, uh, basically, it took uh, some time, maybe like 10 or 15 minutes even. But now I have the ultimate report that has all of the issues and like all of the research juice combined into a singular document that I can use to uh, let's say, fix that issue, or perhaps in your case, to research something very thoroughly. Uh, as you can see here, we have the table of contents with all of the the root causes, the solutions, etc., etc. And this has been absolutely fantastic and instrumental for me uh, to, for example, solve this very particular and very complex issue.
But now there is one more extra step in this workflow so that we bring all of this together. The final step, which is a kind of cherry on top here, is that you put this uh, ultimate research guide now into Notebook LM. So you just create a new notebook, and then you just add this particular source, because of course, in Claude, you can download your source as an .md file, and then you just upload uh, this .md file here. And you can see that over here, we have our full uh, guide uh, related to this uh, particular issue or the thing that we are researching. And the really great thing about it is that now uh, you can ask any question you want about that particular topic, and it will give you the answer based on your ultimate uh, research uh, source. So this is uh, extremely helpful, for example, that when you are encountering uh, different uh, types of challenges, you can now just ask about a specific thing, and it's going to pull from that specific research report.
But that's not all. Uh, you uh, also have the ability to uh, create a mind map uh, from that particular research uh, topic, and you can explore it one by one. So then you can understand um, that issue in more detail. And another cool thing which I have been using quite frequently uh, recently is that you can create the uh, audio overview, which is also interactive. So for example, here you can see that from this particular ultimate report uh, I have created an audio overview that is 23 minutes long, and this is fantastic. For example, if you are trying to solve some kind of issue in your mind, for example, I just use the mobile app for Notebook LM on my Android phone, and then in the meantime, for example, when I'm exercising or just taking a walk or something like this, I can re um, listen to this whole overview. So then already, I'm getting a deeper understanding of a particular uh, subject. Or you can also just listen uh, to it here.
All right. So, the last thing I'm going to leave you with is that you can now put this into action as well. It's not only for research, but for example, if you want to now finally solve this type of issue with Google Analytics 4, you can then just download this whole report as the markdown file, and then you would tackle these issues one by one with a different prompt. So the way I do it is that, for example, I would just now go to uh, Google Gemini 2.5 or something like that, and I would upload this whole um, file as um, full context for that conversation, but then I would say, "Please become a master web developer and analytics expert," and so on. And we are going to be tackling these issues one by one, because uh, that's where you get the power. Like you have all of the context here, but then you need to go through the thing in a more granular fashion uh, with your LLM to finally solve uh, the issue, to make sure that you haven't actually missed anything uh, that is important. We are combining four different tools to basically get the best research possible. Uh, if you like this video, just hit the like button, subscribe to my channel, and perhaps next, you may want to uh, watch another video from my channel uh, about creating really amazing AI presentations with the Gemini app. So, I'm going to see you in the next.