In this tutorial, Shaun Whynacht demonstrates how to use Google Gemini to create a custom "gem"—Gemini's equivalent of a custom GPT—that can generate authentic, first‑person LinkedIn content in his own voice for his agency, Blue Cow Marketing. He starts by launching a new Gemini chat and runs a deep‑research query that pulls publicly available information about himself and his business from web search, Gmail, and Drive. The results are exported as a PDF document, which serves as foundational knowledge for the gem. Shaun then creates a new gem, names it for the intended platform (e.g., "BCM26 LinkedIn"), and uploads the research PDF along with any other relevant files such as video or podcast transcripts. He crafts detailed instruction prompts that enforce style guidelines (first‑person, story‑based, no emojis or hashtags, natural language, LinkedIn best practices) and disables hallucination by requiring direct references to the supplied transcripts. He explains how to configure the gem’s default tool, save, pin, share, and edit it later. Finally, Shaun shows the gem generating eight story‑based LinkedIn posts from a podcast transcript, illustrating how the gem blends the deep‑research knowledge with the new transcript to produce personalized content. Throughout, he offers practical tips—reviewing AI‑generated research, using multiple AI platforms for optimal results, and continuously updating the gem’s knowledge base—to ensure the output remains authentic and useful.
Chapters
So here we are in Google Gemini. This video, we're going to walk through how do we use AI to train AI to create content for us that is authentic to our voice, who we are, and all that stuff. So what we're going to do is we're going to start fresh here, and we're going to create what's called a gem. And that's Gemini's version of like a custom GPT if you're in that ChatGPT space. The reason I'm using Google is I'm a Google business, and we use Google for everything. And while I do use ChatGPT, and I've used it a lot, I'm seeing advancements in the Gemini world of what it can do and keeping in the same ecosystem. So what I'm going to show you here could actually still be done in Gemini, and I'm sure it can be done in Claude and all the other ones that are out there, but it's just the process. Okay, so what I'm going to do first, I'm going to go into a new chat, and we're going to get it to do deep research on me and my business. So instead of me typing, if you watch any other videos that I've done recently, I use the voice VoiceOver feature of Apple by just double-clicking the Control button, then it goes into voice dictation. So I can just say— so you can hear me say what I want to say in this. So I'm going to go in here, I'm going to go deep research, and I'm going to ask it to go ahead and do the research. Now what you can see here is a dropdown for searches— for sources, sorry. You can choose search, Gmail, Drive. Now we can also know here is if you got Google Drive set up and you got documents in there, you can have it search that or your email and research a bunch of stuff, or you can upload files that I wanted to do. But in this case, I'm just going to leave it in search because I want it to pull what's online. I know that I exist online and the information that's there is pretty relevant. So we're going to go in here and say, enable this. I want you to do a deep research into Shaun Whynacht and my company, Blue Cow Marketing. And the reason I want you to do this is because I want to use the information that you obtain to train a custom gem for future content. So research things like testimonials that you find online, our website, anything that's relevant to what we do, who we service, and how we help them. Okay, so I'm gonna do that. And you got the option here for speed. I'm gonna use the fast as well. Um, if you're doing thinking or pro, looks for advanced math, which Fast is all you really need. And we're gonna hit submit. Now this could take a few minutes for this to do its research, so I'm just gonna pause the video and let it go. So just hopping back in before it does, it comes up and gives a summary of what it's gonna do, what it's gonna research, and then it asks you to either edit the plan or hit start research. So we're gonna start it and let it do its thing, do all its little body stuff that is happening. And it will start to go and you can watch it. Hey, if you got nothing else to do, grab a coffee, grab a tea, grab a brewski, whatever, uh, and just see it do its thing. Now the results it's going to give you is going to be quite in-depth. So we see here it's looking at all these different results that are here and what it could find online, and it will find sites for reviews that you didn't know even existed that will come up in there. I do recommend that any of the results you get from AI for the research side, you do scan them and look through them. For the sake of this video, I'm not going to go line by line through that. I'm just going to show you how we utilize this information for future stuff. Okay. Okay. So I'm back. It's been 3 minutes. I went and got myself a coffee while it was doing its thing, came back in and here it is. Here's all the research that it came up with. And you know, you can go through and read through all that stuff, right? I've done this before. So I know the information it's pulling is pretty relevant for what I want. So what I'm going to do here is I'm going to export this as a document. Right. So if you're in the Google Doc ecosystem, it's going to export here. And we'll wait for this to do its thing. So there it is. Okay. So what I'm going to do is I'm going to actually download this as a PDF and save it like that. Okay. Now we're done with this research document here. I'm just going to go back to this. Let's move this out of the way. So many things in the way. Now, if I wanted to, to change this, I could go through and edit the document and whatever, or if it's completely in left field, you can go and, you know, redo it and all that stuff. I will let you know that there are times where I've done this research for clients where I've gotten a better result in the research phase using, let's say, ChatGPT over Gemini, or in these other ones where ChatGPT wasn't as good and Gemini did a better job at it. So if you got multiple platforms, you can do this initial stage, the one that gives you the best results as well. Now, in addition to this, the other thing that I would recommend is if you have a lot of writing or things like that, or you do videos and things, is to transcribe them. Obviously, you're writing, you won't transcribe, but your videos, you transcribe them, or your podcast, to hear how you speak, the words you use, and all that kind of stuff. And I'm going to show you where you would put that. We're not going to go for the sake of time and do that in this. I will do it after for all my videos that I have. The reason for this is we want to give the gem as much knowledge as to how you speak, your content, all that stuff. You got books, upload those if that's the content that you want to utilize in stuff that you're creating. So we're going to go on the left-hand side under gems, and we're going to say explore gems. Now, just like the ChatGPT custom GPT library, there are ones in here that other people have made, but we're going to create a new gem. And we're going to give it a name. So I'm going to just call this, um, Blue Cow Marketing Content 2026. And we're going to describe what it does. So this, this will generate, um, social media content for. Okay. Now. You can create custom— this is what I'd also recommend, creating multiple gems for different things. So I just said for social media, so I'm going to generate social media content and I'm going to say LinkedIn. And the reason for that is a lot of these AI models know how the content needs to be formatted for different things. So if you're doing ones for YouTube and you created one that was just for YouTube, so in this one I'm going to call this— actually, I'm going to change this and shorten this and call this BCM26 LinkedIn, right? It's only going to show my list on this side. I know they got a short amount of space. So I'm going to use that and you could create another one, use the same training document for, let's say YouTube or podcasts or whatever. So you have different gems for different things, right? Different gems for different things, right? Sounds like a really cool song. In one of my videos that I'm also doing this month, I'm going to talk about AI song generation. So now in the instruction side here, this is where we'll start to tell it what we want it to to do. And you may be saying to yourself, whoa, Shaun, what's with the document you created? Where do you use that? Well, scroll down, don't get ahead of me. I'll show you that first, just in case that's driving you nuts. So down here under knowledge, it says add files to gems to reference, get a little plus sign. And we're going to say upload file. And there's that research document. So it's going to upload there. Now if you had other ones, so you did those transcriptions, put them in a document, download them as a PDF or Word doc and add them in here. So the more information you have on here,, the better. All right. The other thing that you could do is you could get your bio and all that stuff and just give it more. And over time, you can always edit these and give it more and more information to, to tune it in. So I'm gonna go in here and I'm gonna give it the instructions. All right. So again, I'm gonna use my voice to text feature of my Apple computer. You will generate LinkedIn content. For Shaun Whynacht's personal LinkedIn feed, period. This will be based off of transcripts of videos he's recorded or transcripts of podcast episodes he's recorded. Now I had a little stumble there, so we're going to get rid of that. We're going to continue on. Double-click Control. All the content you need to create is going to be story-based. I don't want you to hallucinate and go off on a tangent if it is not directly referenced or the concept is referred to in the transcripts in which you have been provided, period, new paragraph. Any of the content that you create needs to be in natural language, easy to read. Don't use large words. Or large paragraph structures. Period. New paragraph. Do not use any emojis in the content, as well as do not use em dashes. Period. All content needs to be written in the first person, as if it was written by Shaun himself Period. As an inspiration, consider the content as if he was giving advice to a really close friend to help them out. Because this content is used for LinkedIn, I give you permission to use some of the best practices you know for LinkedIn content, period. Shaun does not like to use hashtags in the content, so do not put them in the immediate content. If you have a recommendation for a first comment to put on the post, please include that in a separate section. Okay, so that's what we're going to run with there, and that's pretty good. And you go through and fix any of the, you know, stuff that Grammarly wants to jump up in the punctuation, but the punctuation doesn't matter in these instructions. So we're going to go ahead and we hit save. But before I do that, you might say, what's this default tool? This allows you to say, okay, I could create ones that are just for deep research. So you can create tools or gems just for that. You can create them for videos, for creating images, for canvases, or guided learning., but we're just going to use no default tool, which means if we ask it to generate any of those other things or do deep research, it will switch tools for that. So we're going to click save. And now it's created that gem and we could share it with other people on our team, our friends, our family, our grandparents, whatever we want. And now this, this exists and they show up on the left-hand side. Now, if I want to get back into it, I can click on these three dots and click edit, and it's going to take me back into this this editing phase. If it's one you're going to use a lot, as you see, there's only like 3 that show here. You can go here and pin it, which means it's always going to stay up on the left-hand side. And again, you can share it or delete it if you choose to do so. So we're going to click on this and here we are in this prompt. Now I did my podcast episode in Descript. Here it is right there. It's about Agentek marketing automation and we're going to go export transcript and plain text and we're going to go there. We're just going to copy to the clipboard. We'll come back here and I'm going to go ahead and paste it in here. So I'm going to tell it, I want you to create 8 story-based posts for Shaun's LinkedIn based off of things that he talked about in this podcast episode. I'm going to paste the transcript of the podcast below this text for your reference. Use the knowledge that you have been given about the deep research into Shaun and Blue Cow Marketing, if some of the stories or testimonials that you know would be beneficial to the content. And I paste it in there, and we're going to see what comes up here. So without knowing the context, you guys don't know what the podcast was about and what I talked about in there. So it's going to go in and it's going to do what I want it to do, right? So post 1, death of a long email. I have a confession to make. I don't I don't read long emails. I did say that in there, right? If I open an email, I have to scroll multiple times to find the point. I'm out. True. All facts, right? Is automation too linear? Most of us build automation like a straight line. Again, I talked about that. So overcoming that part of will AI generate content that isn't us? Yes, it would if we didn't train it or give it that initial content. So what we're getting here, here is a base knowledge for it to know who we are. As well as feeding it with the transcript at the top. And the key point here is the $2,000 DJ lesson. In that episode, I did not say anything about DJing in there, but I did notice in the research, because it's publicly out there and all the content online, is I did run a DJ business in the late '90s, and my grandmother did give me a $2,000 loan. So here's story-based content. Now it's pulling references from the knowledge that it has about me and building that content. Okay, these are the types of posts that I like to put out there. So this is how we can start to use this to build out, you know, whether it's a custom GPT or a gem or whatever Claude has in there. Maybe it's a Claude Hopper. I don't know. As a guy with size 15 feet, I would— that term is very familiar with me., but this is something that I don't think a lot of people are using properly, right? They're going in and they're just telling it every time stuff about themselves, or they're generating content and they spend a lot of time just reprompting and reprompting and over and over again to try to get it right. But if you take the time to feed it with what you want, think of it like you are hiring somebody and this was your onboarding process. You gave them the, you know, this is who we are, this is what we do, all that kind of stuff. Then great, perfect. Now, That's what the AI will use. Now, if you look at this, you're like, okay, I always want to offer, you know, let's say the free book that I have. You could tell it in the instructions at the end of every post, offer, you know, a link to my free book. Here's the URL, right? Go that way and do that. So let me know if you have any questions about this as well, because I'd love to answer those for you and help you out.
