#680 - AI Toolkit for Amazon Sellers

Bradley Sutton: Today, we've got one of the top minds in the whole world when it comes to ChatGPT, custom GPTs, Rufus and more. And on this show, he's even going to give a special link of some of the private Amazon custom GPTs that he has exclusively for our Serious Sellers Podcast audience. How cool is that? Pretty cool, I think. Hello, everybody, and welcome to another episode of the Serious Sellers Podcast by Helium 10. I'm your host, Bradley Sutton, and this is a show that's completely BS-free, unscripted and unrehearsed, organic conversation about serious strategies for serious sellers of any level in the e-commerce world. And for the first time in our podcast, he's been on the AM, PM podcast before. He's been on webinars with us before, but the first time on the Serious Sellers Podcast, we've got Andrew Bell coming live and direct from, where were you even at right now? Andrew Bell: Owensboro, Kentucky. Bradley Sutton: From Kentucky, live and direct from Kentucky, that's right. We had some running jokes in our elite workshop. There were some strange keywords that you had never heard of and your go-to excuse was you're from Kentucky. But I think you're selling the, wait, Kentucky is like, there's the blankety blank state, there's the blank state, Kentucky, bluegrass state, right? Andrew Bell: That's right, yeah. Bradley Sutton: So you're selling the bluegrass, your bluegrass compatriots a little short there. But anyways, welcome to the show. You have been in the e-commerce world for a little while, but let's go before e-commerce. We talked a little bit about this on the webinar, but people on the podcast might be the first time listening to you. Were you born and raised there in Kentucky? Andrew Bell: Just across the river, so yeah, basically. I was actually from a small town called Santa Claus, Indiana, which is real. It's not fake. It's a Christmas-themed town, live on Blue Shoes Drive. There's a theme park called Holiday World, but Owensboro is just about 30 minutes away, so just across the river. Bradley Sutton: Now, I know sports, that direction is very popular. So is everybody like Kentucky Wildcats fan out there? Andrew Bell: Yeah, Kentucky fans and my wife's brother runs at Louisville. He does like the 800-meter and stuff and so I have a Louisville like a plate on the back of my car and I'm pretty much… Louisville Cardinals, right? Bradley Sutton: Is it Cardinals? Andrew Bell: Yeah, that's right. Louisville Cardinals and let's just say I'm not very welcome in Owensboro for that. Bradley Sutton: Yeah, in certain places, people understand this kind of thing in the UK for like soccer teams where if you go to a neighborhood that's all Manchester City, you better not be bringing your Manchester United stuff there. In Alabama, if you're wearing your Auburn jerseys, you could get beat up in some Alabama hotbeds. I think here in California, people don't care too much. There's a lot of sports teams out here. We have some rivalries like Padres versus Dodgers and things like that. We're actually going to be talking about rivalries, like controversy about what is happening with AI and stuff, so this kind of goes to it. Back to your backstory, you did not go to neither Louisville nor Kentucky for university. Where did you go? Andrew Bell: I went to Moody Bible Institute, which is a Bible college, a place strictly for preaching. They don't even have business majors there. It's strictly for preaching, missionaries, and then I, of course, studied ancient Greek and preaching. I even to this day, I'll read in my Greek New Testament for my devotionals too. So I like keeping up with it. Think of it as like my Sudoku for the day. Bradley Sutton: How in the world Do you go from majoring in ancient Greek and biblical studies to e-commerce? Well, fill in the gap there. Andrew Bell: Oh, yeah, obviously. It's like, yeah, no, it's a natural transition, right? Bradley Sutton: Yeah, of course. Natural segue. Andrew Bell: Yeah, exactly. But no, I basically what I did is I had come home from school, you know, graduation and such and met my wife and Pretty awesome. Started working for a church and then I decided, I don't know if I want to do this anymore. I think I'm going to go into e-commerce world. So there was this company at the time called Touch of Glass that wasn't really doing Amazon. I thought, oh man, this is a goldmine for Amazon for products to go up on there and to sell. They're very niche. They're always the highest priced items. I believe there was a lot of like, you know, demand for it as well and so that's when I enrolled in. This is not a promotion necessarily but enrolled in Freedom Ticket. So you're my first teacher. Bradley Sutton: So that's pretty. Yeah, Kevin. Andrew Bell: Yeah, you and Kevin. Yeah, yeah, that's right. Bradley Sutton: Well, I'm wearing, we used to have an old Freedom Ticket. We had the original Project X and I'm wearing a, I think you can see it here, a hat that has an X on it. That's why I'm wearing it. This is not just a Mighty Ducks hat but this was the reason. I get hats sometimes if it has like tie-ins to Helium 10. We're going to talk a little bit about how to sell on Amazon FBA & Walmart. A large catalog of products they just weren't really optimized or weren't even selling really well on Amazon and so you came in and what were you able to do for them? Andrew Bell: I was basically able to take all their ASINs, so about 4,000 ASINs, and I basically started everything from scratch. I'm talking pickpacking, shipping, doing FBA, to learning everything on Seller Central, to building a brand store that eventually made a million dollars. Out of the 7 million we made on Amazon annually, within four years, I was able to get us there. One of the things that leads into this as well is Amazon listing optimization. I went through Helium 10 and did all the training. I was the guy that was doing Cerebro. times on every product and that's just probably a minimum of what I would do. How long did that take? Bradley Sutton: That's a lot of products. I'm assuming you're doing keyword research on all these products. You're editing, you're making the listing, maybe commissioning some other graphics. Is this a project that went on for a couple of years? Andrew Bell: Oh, yeah. I mean, hours, you know, 40, 50 hours a week just spent on things like that with listing optimization. But nothing was more fun than that. I would just sit back and go to a Starbucks and just literally do optimization and just have a lot of fun. One day, I remember I was doing Tiffany-style table lamps. And someone asked me what I was doing. I was like, oh, I'm doing Tiffany-style table lamps. And I was kind of showing what was going on. By the end of that day, I had 28 Tiffany-style table lamps done. For optimizations, and I kind of scaled the strategy for myself where we use magnet, use Cerebro, built keyword lists for every single product. So I had a unique keyword list for every product. Of course, I would reuse general things like metal wall art, let's say, and, you know, wall decor. And then if I was doing accent tables, even though I had a vintage accent table, I would still use the accent table list. So I had varying degrees of lists as well that I would apply, and this is back when I would just use Scribbles. This is before Listing Builder, if you remember Scribbles. I love using Scribbles. Scribbles could show you what keywords were used and what weren't. It doesn't have quite the same sophistication as Listing Builder, but nonetheless, I really like that. Bradley Sutton: It's an oldie but goodie. Andrew Bell: Yeah, it is. It's tried and true. That's right. Bradley Sutton: Now, were these Amazon lists, like these products, all 4,000 of these, did they exist on Amazon but they were just like thrown up there or they weren't even on Amazon at all? They were just on the company's website? Andrew Bell: Yeah, they weren't necessarily on Amazon. Bradley Sutton: So you were basically doing it from scratch. It was like somebody who makes a brand new product. It was almost like you were doing 4,000 product launches then as it were. Andrew Bell: Yeah, absolutely. Yeah. So if like we had a ton of Tiffany's style table lamps, those are the ones that I would optimize. And that was kind of like a one man team. And of course, I had departments at my disposal as well because this was a very seasoned company as well. And everything we did was our exclusive. So we had no competition whatsoever in the buy box. So we had to compete with other products on price. And it came down for me, it came down to like, We have to show up over those people because we have no other option because they're going to go with the other product because it's much higher priced. Bradley Sutton: You're talking about competing products or distributors of the actual brand product that are on your piggyback and on your listing? Andrew Bell: No, no, no. So we did not have to be on BuyBox. We had 100% BuyBox because they're exclusives. Only ones in our listings, right. I'm talking more people within our niche competitors that we had. So per category, I would track about 10 to 15 different competitors. So for example, with Metal Wall Art, we have the largest collection of metal wall art possibly on the planet. I know we do on Amazon for sure. And everything is handcrafted steel, all that stuff and all sorts of different themes. And so I track all the competitors related to that. And we were always the highest price item in there. But one thing I noticed is our keyword research was always better. And that made us stand out over others. And that's why we still sold. Bradley Sutton: Yeah, that I mean, sometimes people are like, Oh, no, I don't want to get into this niche. Everybody's so low price. Well, it's not always a matter of playing the price game. You know, there's some niches where you actually play the price game or playing the price game means opposite of what you would think it does. It means like you should be the highest price, you know, like people in the baby category. There's a big niche for like, There's parents out there like, ìNo, I want the best for my newborn. I'm going to go buy some cheap. I don't want to be that parent that buys the $8 stroller. Give me that $80 one.î Amazon success is not always about fighting Chinese sellers as some people think or sellers from factories in India or wherever where they're barely breaking even. You don't have to play that game. I think another takeaway is For all you listeners out there, maybe you guys have been long-term listeners of the podcast and for whatever reason you haven't started selling on Amazon yourself, Andrew's path is very similar to my path. Actually, before I started selling on Amazon, I was a consultant for others. So if you take in a lot of knowledge even without selling your own product, that doesn't mean the only application of what you guys have learned in Freedom Ticket or this podcast or other places can only be applied to your own brand. Become a consultant. Like Andrew made a career out of helping other companies and even getting employment for putting it into practice without ever having sold his own product. So that's a very, very viable path. Thinking about selling on TikTok shop or maybe you are already in it and you're ready to scale? Unlock all of Helium 10's brand new TikTok shop tools with our diamond plan. Everything from bulk Amazon to TikTok listing conversions to instant Amazon MCF fulfillment. Best of all, you use the code TT10 to get 10% off Diamond for six months, even if you've used a coupon before. So go ahead and upgrade and let Helium 10 do all the heavy lifting for you so you can focus on what really matters. For more info on our new TikTok shop offerings, visit h10.me forward slash TikTok. I'll see you there. Now, after, you know, being all in the game, as it were, at what point did you start getting obsessed with all things AI? Andrew Bell: Well, as soon as my wife actually introduced me to ChatGPT. And she had been talking about it. She's a content creator. This would have been November 2022 when it launched. And she was like right on it. And I'm like, oh my gosh, it seems like she's always ahead of me, teaching me things, you know. But yeah, I basically the message limits just went up and up. And up, and I just kept, you know, hitting those limits every single time. Once new models and such came out, you know, what ChatGPT 4.0, you know, or ChatGPT 4.0 cost back, you know, 2023, I believe April when it came out is so cheap now. In fact, I don't think they even have it. They've retired regular GBT 4.0, at least on the, when you go and you have the regular account and such, and they have ChatGPT 4.0 instead. Okay, and so yeah, yeah, so I definitely got interested doing that. And so then that led me to think, oh my gosh, how can I apply this to my job? How can I scale already a robust strategy that I have for listing optimization with keyword research, marry the two together? And create a prompting philosophy that would back all of this. And this was well before... Bradley Sutton: You're probably wishing you had it before you started those 4,000 listings. Andrew Bell: But because of that, I was able to develop a really good prompting philosophy. I was able to use those SOPs and such and put them into a type of logic that would scale listings for other sellers in a way that was both keyword rich, but also narrative style, like driven. So like built to convert as well. And that's the success that I had saw before and so I did that and married also too with best practices. And this was well before GBTs came out. So then I decided, you know, I'm going to, you know, make a GBT out of this a year later. And that's, you know, by then I'd already like fine-tuned all the prompts across different models. So when GBTs came out, the GBT store, it made perfect sense for me to turn everything into GBTs. And before you know it, and to this day, I have about 350 GBTs. I actually counted today around 350 and 30 of which are Amazon related. And now I have the top. Bradley Sutton: When I look at 30 out of 300 are Amazon related. Andrew Bell: Yeah, only well, if explicitly Amazon Yeah, all the others are like for fun stuff. Bradley Sutton: Like for example, the I think I told you about the poetry GBT about the poetry because you got the number one poetry custom GPT and right? Andrew Bell: Yeah. Yeah, it's actually funny to actually have the number one you wouldn't think this number one pregnancy GBT to You're thinking, wow, well, what does it do? Well, I had I had twin girls. And so I coordinated when I was you know, we and so we had to go into the office frequently for her to get like looked at and stuff even more because if you're twins, it's high risk pregnancy. Anyways, I coordinated with all the obese there. And like shared my passion for AI and stuff and like got their consultation, you know, out of like the say the 10 times that I was there, I really made use of that. Because when you have OBGYNs in like more critical care unit type stuff, you get I think like the best ones. So I was able to use that, put them into like a really good knowledge base, like use what they had and put in a really good knowledge base and create a really good GBT, which is actually my highest rated GBT, 4.8 out of 5 stars for 200 ratings and reviews. So it's backed up by a significant amount of data, which is cool. So I did that. I had a passion kind of doing it for everything, you know. Bradley Sutton: Okay. Well, while we're on this topic, There are some people who might have misconceptions about AI or they're new to it, they don't know how it works, and they might have some initial negative reactions such as, oh, I tried it for the first time. I ask it about me and it made up some stuff or maybe somebody has seen asking ChatGPT about Amazon SEO and it's making up nonsense that literally doesn't exist like the Amazon A10 algorithm and stuff like that. So how can I trust AI and stuff? This is my Amazon listings. How in the world can I trust my business We're going to talk a little bit, you're an expert on how to use AI in a way that lessens the number of mistakes, hallucinations, whatever you want to call it. But in the first place, why do things like that happen? Why is Amazon, or why is Amazon, why is ChatGPT talking about an A10 algorithm that literally doesn't exist? I know the answer to this, but there might be people out there that don't know, so you can probably explain it better than me. Andrew Bell: It's because when the AI is trained over time, it's trained on The articles that have been written at the time and a 10 was a way to use to boost themselves in the search ranking because everybody was searching for it because it sounds really good like oh the next model we're going to type in a 10. And so tons of articles are coming out well by the time recently what you know when that happened. It was already trained on all those articles that said A10. So AI really thinks that A10 is like the most important thing. And even when you search for it, it'll actually default to that until you press it. You press it and say, hey, I don't think that's correct. And then it'll say, well, I think it is. I'm like, you need to search a little deeper. And then finally it corrects itself. With that so and that but that has a lot to do with like just having bad data to like bad background information and kind of banking off like the hallucinations existed before but like with with keyword. With the prompting philosophy i had that i developed because it was already so robust and really good. I was able to then, like you said, reduce the number of hallucinations that exist. But when I was using Helium 10, that gave me the trust that I needed to go forward and create the best listings possible. So what you shouldn't do when you're looking at my GBTs is say, I don't need Helium 10. No, if anything, you need them more than ever, right? Bradley Sutton: Let's talk about that a little bit, not just about Helium 10, but if somebody just use your free, and by the way, everything that Andrew is talking about today, poetry, GBTs, I guess the pregnancy one, all this stuff is free. He does this for fun and out of kindness of his heart, but if you went and found one of Andrew's 30 Amazon GPTs, whether it's listing optimization or whatever, and don't connect it to anything, it's not terrible. It's going to give some decent information. It can write a decent listing, but why wouldn't it be fully optimized? First of all, Helium 10's database is not open to the world, let alone GPT or open AI. What is ChatGPT basing a listing on? Because it's obviously going to be based on keywords no matter what keywords make up a listing. Since it's not connected to Helium 10 and if somebody doesn't put the right keywords in, where is it even getting the keywords? Is it like Google? Is it just like common sense or what? Andrew Bell: Yeah, I think it's a mixture of both. If you don't have it, use the tool of search. Typically, it'll just kind of like you said, make it up on its own. You can tell when it's Like kind of bull crapping you guys are like, no, that's that's not true. But who's you know, if you don't know better, it's like and you're just starting on Amazon, you think oh, this is good. And even if you search it's giving you a top 10 keywords for Google and then it uses ones from Google's top 10 from Walmart, bringing Walmart. Oh, these are the top 10 from Amazon, but it's not even in your category. So it's like it doesn't have enough intelligence to know that because when it's writing it, It doesn't have your proprietary data, for example, or Amazon proprietary data to back it up. So it has to rely on something else. And because the number one policy really for ChatGPT for OpenAI is to assist the user and give something useful to the user no matter what, it'll go behind Like behind your back and make you think very confidently that these are the keywords that are going to help you show up. Bradley Sutton: Yeah. So like, you know, again, there's, I'm not going to try to BS people out there, but unless Helium 10, which we're not going to do, opens up our entire billion-dollar database to the world, including AI, or unless Amazon all of a sudden makes, which it's not going to, makes search query performance available to be crawled by AI and opens up brand analytics to everybody, what are you always going to need to pair with the use of AI? Whether we're talking product research, whether we're talking keyword research, whether we're talking listing optimization, what is the limitations of AI in this world that we live in where A lot of data sources, be it from Amazon or other places, are gated. Andrew Bell: Data is limited to the amount of information someone's willing to disclose through their own source that's not gated, right? That no paywall exists. And that kind of depends. But if you're willing to provide the information from Helium 10, plug in the keywords and have it trained on that, and then it memorizes that, that's a good thing. However, you don't want to rely strictly on their memory because sometimes they'll hallucinate with multiple things. It can be worse for hallucination. They start combining hallucinations from before with real data. And then make wrong hallucinating interpretations and whatnot. So when you marry the data keyword research that like you've done through or let's say this a workflow because it's not just about you asked about like what you should be doing if you don't want AI to hallucinate as much. You need to be mimicking real-world workflows that actually produce something that's good because otherwise AI is going to assume a framework that might not even be relevant to your niche, right? And what you need is you need a workflow that's best practices from the industry because Even though Helium 10, let's say, let's say you're not paying for Helium 10, let's say you're not paying for another one, even though that they'd still come out with articles that talk about like how to do keyword research nonetheless and applying those methods is a significant way to reduce the number of hallucinations and to give you a better answer. But what makes it even better is having the actual data behind the framework itself. And the workflow there, that's where you get the holy grail of all optimizations is when you have a superior prompting logic behind superior data going forward. And in this case, we're talking keyword research, product research, and the like. Bradley Sutton: It's like the old adage, garbage in, garbage out, and then whatever the opposite of that is, is AI can only work with what it can work with. It does a darn good job even if you don't give it anything. It's better than just, hey, let me ask Google something that it has no idea about. It has so much more data in it than what people are used to with using Google and stuff, but it's limited with the ability. There's nothing wrong with doing product research on there for like, hey, what's trending out there? But is that going to give you the best Amazon knowledge? No, like here, I just did. Let me share my screen here. I just did an example right when you were talking here about I asked ChatGPT, what are the top selling wooden egg trays on Amazon, right? So it gave just some random like answers here. And then I clicked on each of these links. And first of all, it's not even showing me what an egg tray is. These are like the stackable egg racks. I also sell one of these. And these aren't necessarily the top selling ones. Interestingly enough, it says, based on customer reviews and popularity, like, well, it would not know anything about popularity, you know, per se. But here's the search for wooden egg tray. A lot of those don't even come up here. And our product, here's one of the Project X products right here, is one of the top selling ones. And it's nowhere on there. Now, what could have changed this is if I had Downloaded from maybe Amazon Product Opportunity Explorer, some click data of all the stuff that's a wooden egg tray. If we're talking Helium 10, I could have gone to BlackBox, typed in wooden egg tray, downloaded the CSV file that has all the sales. And now if I say the same exact question, I probably could do that right now, but I'd probably take five minutes, so I'm not going to do it. But if I say, what are the top-selling wooden egg trays on Amazon, here's the file of the data of what's going on, it would tell me exactly the top five ones that I'm looking for. So again, it has to do with the data that you put in and personally, I don't think there's ever going to be a world where it's just going to be able to function on its own when we're talking about different marketplaces that have proprietary information. What's the best ones from Shopify or Google? Who knows? Maybe that one. There's public domain information about what's the most searched pages and maybe how many pages are viewed. It's not as private as like Amazon or Walmart.com or TikTok shop. But guys, stop thinking you could just replace, oh, I don't need search query performance anymore. I've got ChatGPT. I don't need Helium 10. I have ChatGPT. There's never going to be a world where you don't need to augment it with data. Now, along the lines of the hallucinations in general, maybe if you can share your screen. And if you can give, I will try and describe this as much as we can for those who are listening like on their morning run to work or something. I hope you're not running to work. Your morning jog before work or we'll try to verbally describe it for you. But can you give some examples and tips on how people can use ChatGPT in general or with Amazon related tasks that will help give them better results because this is something that you are super good at. Andrew Bell: I think one thing I want to show is I can screen share here. Product Research kind of shows you the market and if it's worth putting a product up to begin with. We did this in one of the others, Bradley. Bradley Sutton: Which version of GPT are we looking at here? Andrew Bell: So this is the Product Research GPT and it's actually functioning off O3 reasoning model, which that means it spends a significant amount of time basically thinking through every different facet. So what this is going to do is it's going to go through 10 different possible dimensions in the research process for product research. And within each of those, it has its own research prompt. So it's not just, hey, find me the best product, right? The next best hit. It's like something you can put in your own product or a product that you want to try on Amazon or possibly launch. And it will produce for you a product research that's based on 10 different research prompts across 10 dimensions, each of which provides its own search as well. Bradley Sutton: So first of all, my first takeaway is one way you can get better with just using ChatGPT is actually. Using a custom GPT like you've created that is doing like 75 million, not that much, but a lot of prompts that you've programmed so that somebody doesn't have to go and type out 17 paragraphs of instructions. You've kind of like automated it here to go that extra mile. Andrew Bell: Exactly. It saves you the work. You don't have to store a prompt somewhere either, which is pretty nice, I think, as well. And you can then plug in whatever product you want and it keeps you in context no matter what. And that's the beauty of a GPT. Always keeps you in context. You don't have to store a prompt anywhere else. And it's extremely reliable as well. So here with the prompt, let's talk about, I'm going to kind of talk you through what's happening with like the reasoning process behind it. Okay. What product you want to do? Bradley Sutton: I was thinking, let's just do the, what about the one I just showed? Like a wooden egg tray or something. Andrew Bell: Yeah. Bradley Sutton: Okay. Andrew Bell: So I'm going to type in wooden egg tray. And basically, it's going to go through its thinking process. And this is the distinction between a reasoning model and non is it it doesn't like think through everything all the way it'll produce an answer that is immediate with a different, you know, with the regular model, but with a reasoning one is not the case. Bradley Sutton: I like how it's kind of thinking out loud as it were. It's like showing you. Andrew Bell: Exactly. It's showing you the work. Bradley Sutton: That's different than, yeah. Andrew Bell: And that fosters trust in my mind of what's happening. So you need to follow and track it all the way. So it's like the user mentioned this, but there's no real direct question attached. So then it thinks through, it says, okay, for each dimension, I'll create two to three search queries, run them through a web search task. Each search query will target specific aspects of the product. I'll proceed with the structure for a thorough investigation. And then it's talking about, okay, I got the three searches here, as you see, the first search, and then it goes through five different sources, domains. So it goes through trends.google, DataBridge, Market Research, Amazon, Spur Research, RuleHandmade.com, which would make sense to search something like that. So niche, because we're talking about... So then it keeps going and does the same thing. And then once it retrieves the results, it summarizes the findings, gathers the top sources with citations. And then it says, I'll do this for each dimension. So it's repeating the same step consistently, you'll notice here. So here you have three search queries, about the same number of domains. And then as you continue going down, three search queries, same number of domains. And then you go down again. Bradley Sutton: So many steps here. This is amazing. Andrew Bell: Yes. And so it goes very thoroughly. And that's the thing using reasoning models. I highly encourage it because it cultivates and fosters trust with the model. And that's that's what you need. Because otherwise, it's like you're you're kind of trusting it in the dark, especially when you're talking about data, which is, by the way, the number one reason why people are unwilling to use generative AI with PPC, as an example. Bradley Sutton: Yep. Yep. Andrew Bell: Because you're trusting with your money. So now it's looking at competitors. And so it'll go through and look at wooden egg trays in different retail channels. So for Amazon, Walmart, Etsy, wholesale suppliers. So it's searching Etsy here, bear.com, Alibaba, backyard, barnyard. I mean, these obviously sounds like something that it would go into. Then dives into several other areas, competitor brands, pricing strategies, distribution channels. So you see the consistency with which it's going. It hasn't stopped the consistency. Bradley Sutton: Now how much, how, what did you have to do that this custom GBT is going into so much depth with just somebody typing in wooden egg tray? Like, like how much, how many lines of prompts did you even give this? Or is most of this just, it's using its own common sense of what it thinks it should do? Andrew Bell: No, this is something that has to be intricately put in because otherwise we could take it a million other directions. Bradley Sutton: Yeah, I was about to say like, this is pretty good for making it up on its own. Okay. Andrew Bell: Yeah, absolutely. So yeah, as you keep going, then it goes in, I'm digging into future trends for the wooden egg trays, similar products, search for insights on these various things. And then it keeps going. And at that point, it's gone through all the dimensions. So it faithfully went through all requested dimensions for the product idea of Wooden Egg Tray. So each dimension now is presented with a research prompt that was used. The web search query is actually executed. So it shows its work here even more. Concise insight summary. And then the key findings. So here we find out the general egg tray market, all materials is valued at $18.8 billion at 2025. Yeah. What is a small but growing premium sub-segment? Keyword tools show 3K, 5K average monthly US searches for wooden egg trays. Steady interest but not yet saturated. Google trends indicate mild but consistent five-year growth for wooden egg tray peaking each spring. But even here, look, it cited this and I'm sure it's citing from another source, but like notice it's just the Amazon page that it shows. Bradley Sutton: So this is another thing where like having even better data like, yes, sure, this helps because obviously, yeah, would an egg holder listing does not say in there, hey, this is the search volume is this. That's not exactly on Amazon. So yeah, that's why people have to be careful about, you know, no matter how good the prompt is, it still can, you know, like once it's kind of like, that seven-year-old kid when there's guests coming to the house and they want to kind of like show off and look what I can do, you know, it's like, hey, I don't want to disappoint, you know, so let me, you know, show off a little stuff here. But then what I would do right here is I'd be like, I would have maybe, if we're not talking about products and we're just talking about the general market, I would have maybe gone to historical Cerebro and downloaded like maybe a few of the months his or Magnet actually, Magnet would be better in this case. Any keyword that has egg tray in it and then some of that historical search volume. And if I were to manually like download in Excel all these things from Helium 10 and find out which keywords are increasing in search volume over time or decreasing, that would take forever if it's even possible. But I can just literally just download it and give the raw data and then now all of a sudden this prompt would probably show me even better data based on what Helium 10 is giving. Absolutely. Andrew Bell: Way better data, I would argue. Yeah, absolutely. And then as you go down, Competitor Landscape says, markets fragmented boutique Etsy makers, DTC farmhouse brands, large import resellers on Amazon, eBay. No dominant national brand. Differentiation comes from wood species. Kaya, Walnut, Bamboo, Capacity and Aesthetic. Ceramic metal shelter holders compete on price. And then you have the customer pain points. So surface negative reviews, complaints, and it gives the, again, this is important because it gives the research prompt and the key findings are common complaints, shallow holes that let eggs roll, rough machine machining that splinters, finishes that aren't truly food safe, and trays too large for standard fridges as well. It looks like it's taken from forums as well. So like, you know, reputable forums have come before. And you notice too, it'll take a little bit from Reddit as well. And here's the thing with this product research tool, you can actually be able to fine tune it with your sources. So there's a new version that I'm going to have coming out where you can actually be able to use whatever sources you think are most reputable for your product research. And imagine again, having like putting your Helium 10 data into a product research prompt like this. Bradley Sutton: We're going through pricing strategy, elasticity, and remember guys, those of you watching on YouTube, all he did was type in wooden egg tray. He didn't type in a question or anything. Now, you can't go to ChatGPT, guys, and type in wooden egg tray and expect this. The reason, again, this is coming out is because this is one of his 30 chat or custom GPTs. Real quick, for the people who are taking notes, where can they go to see this exact GPT to use it and your others? How can they find it on the interwebs out there? Andrew Bell: Well, you're going to get a link too. We're going to share the link with you. Bradley Sutton: In the description. We'll put that in the description. Perfect. Andrew Bell: Yeah, in the description. Yeah, I would say that's best. Bradley Sutton: And then if for some reason I can't get that or I don't remember it's there, if I type in Andrew Bell Custom GPT's Amazon or something, would that probably take me to the right place? Andrew Bell: Not this one because it's been gated for the webinars and workshops. Bradley Sutton: It's a special one that you're giving freebies that not even everybody else has access to. Wow, okay. Look at that guys. What happens when you listen to the podcast? You get special freebies that nobody else can get. I love it. Andrew Bell: Yeah, absolutely. You'll get a couple too. There's like two or three that are not open to the public yet, in particular, and not even distributed through many other channels other than Helium 10, for example. So yeah, you'll get this product research one for free. In fact, I have several custom GPTs you can use. I have one, two that I haven't released, product research, product demand ideas, And e-commerce script agent and then a special one called no em dashes, which if you've ever heard of AI, you've probably heard, Oh, having dashes is a, you know, a sign of AI generated content. So if that's something you don't like, it's like, okay, you just plug it into no em dashes and it'll, it'll take care of it. Bradley Sutton: Let's switch gears a little bit. I mean, still talk about Amazon, but another thing you talk extensively about is new things that have to do with AI that actually pop up in like Amazon search results and probably the most notable one would be Rufus. Now, I've done some studies where you can kind of see that in search at least or in my opinion, for me personally, I don't use Rufus that much, but when I do, it's on the product page because I'm like trying to like Hey, give me a summary of the reviews or give me the price history and then when I did a poll out there It still seems like that's the predominant because that actually that's in my opinion That's where it's most useful is on the product page to do stuff that just the human eye can't do but still Rufus is coming up in like, you know search bar and search results and yeah, there's many ways so so it's something that you know, I You know, shouldn't be the very first thing I think that people optimize for, but anything that comes up in Amazon that affects how your product could be found, even if it's only like one out of 100 customers or right, we'll use it. It doesn't matter guys. You need to be optimized for this thing. So in your opinion, What are the top things that sellers today should be optimizing for in today's world? Obviously, we can have a podcast next year. I'm sure the answer is going to be different. But right now in 2025, how should sellers be optimizing their listings when it comes to for better visibility like with Rufus? Andrew Bell: Absolutely. Well, there's a lot of things you can do, you know, you know, techniques of where, you know, you ask all sorts of questions about your product and wherever it says, well, the product information doesn't provide this, but the customer reviews say this, you can basically take that information, that question. It's like, okay, if that's not answered somehow in the bullet points, here's, here's a big thing. Rufus takes from everything on your product detail page from your title, To your product description to your attributes, bullet points, no matter how long they are, Rufus will take it. Rufus basically sees your product page as a knowledge base. And takes from that specifically when it's on there. Now what's pretty cool is when you use Rufus, let's say on the homepage, it'll provide you much more general things. But then when you search with, go to the search page and then ask questions from the search results page, you're gonna get something deep within for that search term. So for example, if I'm typing in metal wall art, you're gonna see options like, hey, compare these metal wall art to one another. But then when you go into a specific product, it actually mines down even deeper. And you can see things like show price history, You can see customer reviews with images that are provided too as well. There's a new feature actually that you can set the price of what you want. Like if it's 5% off and that's what you want to see, you can set it now and Rufus will keep an eye on price for you and give you a notification when a price on a particular product has changed whether it's 5% or 10%. Bradley Sutton: What does that notification come through on? Is it just like your Amazon app or what? Andrew Bell: Yeah, it's like on Yeah, on the Amazon app. Bradley Sutton: And so yeah, but for deals like Prime Day, you know, depending on when this comes out, that's like, maybe I need to buy something. I'm like, you know what, I'm gonna wait a couple days because it's gonna be Prime Day or Black Friday. Let me set a notification. Like, hey, let me know if this, if this product, I didn't know you could do that. It's pretty cool. Andrew Bell: Yeah, but it only works on the mobile app right now for that in specific. And then one thing too is Rufus, I would definitely do, it's called visual label tagging. So putting text on your images because Rufus can actually read and take into factor. In fact, this is pretty cool. When Rufus is asked a question, it'll take not just in all the text, but it'll take every image, look at every image at once. Find the ones that's most relevant, right? And the most relevant one is going to be the one that answers the question with the text or implicitly does it and then chooses an image based on that. And if it's a certain score above, I think it's like points. It's 70% likelihood. I think it'll like actually bring that image into the chat. So having text on your image helps you do that. And what's really cool is Rufus is actually seeing the image itself too. So one time when I typed in for a rug, I said, Hey, I want to see this rug next to a chair and a couch because I saw an image of that and it actually brought that image in. Now it took four tries, but that's because it's deciding how relevant it is to bring it in. So when you have both the image and text that answer the question, it actually helps you bring it in. So if someone's asking, what does this look like up close? Don't just put, you know, this is, you know, here's what it looks like up close. Here's the different, you know, fibers, let's say, of a rug, but actually show the image of it up close and it'll take both into factor, the text and the image, giving you a higher likelihood to bring that in the chat, which would then lead, I think, to better conversions because then, you know, the person's seeing it from a visual point of view as opposed to not. Bradley Sutton: Yeah, okay. That's good to know. Yeah, I've given examples, I think on this podcast, or at least on stage before where, um, One of the things mentioned is making sure that the questions are answered in the correct way and in a positive way for your company and your own listings, right? And if not, you go in and change your listing. And so those of us who have been doing listing optimization on Amazon for years know that sometimes, hey, I want to be indexed for a new keyword. You edit your listing, you put the keywords in. Sometimes it takes like at fastest, maybe like three, four hours to be indexed, maybe a day or two. Guys, when I did something for Rufus, There's a question that was not being answered correctly. No, I'm sorry. It was being answered correctly, but my listing didn't have the right answer. It was something like, hey, is this coffin shelf durable or something like that? And it said, well, the listing doesn't say, but the reviews say, no. I was like, oh, crap. I don't want this. I don't want this to be the number one thing. And so I literally changed one of my bullet points to say, this coffin shelf is very durable because of its high-quality wood finish. Guys, 10 minutes later, I asked Rufus the same question, or it was the auto-complete question. It said, yeah, you know what? According to this listing, it's very durable because of this and that. So it is super fast. So it's something super easy to do, guys. There's only like so many auto questions that Rufus has in the autocomplete. You can like optimize for like less than five minutes here. So it's something that you definitely should do. Now, as far as other ways of influencing Rufus, like to me, like something in the future that would be, or even now if it's possible, I don't know, I've never tried it, but is being able to influence what are the autocomplete things that come up like in the search results, because that's where it's kind of like wide open. Somebody who goes on your listing, they already must have some buyer intent, but have you found any way to kind of like Change what those auto-suggest Rufus questions are either on a page or on a search results. Andrew Bell: No, there's no known ability or way to to do that now that being because okay, so that being said Rufus runs on according to its patent is run on click training data. In other words when a questions clicked that's immediately recorded When it's on a product page, so for example, if I'm on a metal wall art piece, let's say it's abstract, et cetera, and I'm asked a question and I very frequently ask about the size of the item, that click training data with that product will be taken into factor. So yes, that type of question might not show up when I'm looking at a different type of product, let's say for a coffin shelf. But if I go to another wall decor piece, it's going to make the assumption based on that click training data that I'm going to want to ask about size and then it's going to put size first. Now the validity of that we are not 100% sure it's coming up. However, there's evidence recently just came out. I just posted about it recently on LinkedIn where The customer questions that people are asking on the search results page are actually becoming a frequently asked section bar on the search result page on Amazon. So it says customers also ask and it's the Rufus questions there. So the click training data has actually paid off because and it's showing you the most popular ones that are being clicked for Rufus. And so that's actually showing up now across a number of different categories and your brand. I've noticed across almost every brand has had that. So if you guys have a brand, type in your brand into Amazon. Right now this works only on the mobile app and scroll down just a little bit and you'll see customers ask and it'll like give questions and answer questions about that brand. And those are answers that likely do come from Rufus because of that click training data. Bradley Sutton: Okay. So guys, I mean, this is still, I mean, I know it's been around for like over a year now, but it's still, that's infancy when it comes to Amazon, you know, so, so, and it's changing and, and you could see tests that are happening every day where somebody on LinkedIn will say, Hey, look, what's showing up in my browser and nobody else has that. Well, what does that mean? That means Amazon is doing some kind of test because they do a lot of tests with new and emerging technologies. And so, so guys, this is something that Amazon is going to continue to iterate on. And right now there's nothing from like, the Amazon API or brand analytics that gives you data, but hopefully fingers crossed in the future. We're going to have some better data from Amazon that helps us to optimize even more and helps us to understand what we need to do. But in the meantime, guys, it's super simple what Andrew has talked about that you guys need to be doing now. But trust me, guys, this is going to affect later on how you might do your advertising and things like that. I personally don't think it's ever going to take over the shopping experience. And the reason why is because Amazon is like if I go to Amazon, and because I'm looking for a, and I literally bought this last week, a 64 ounce insulated water bottle. There is nothing that can ever be made other than like a mind reader that is going to beat the experience of me typing in to search 64 O. Z. Insulated water bottle and then seeing the results. I am not going to sit there and have a three minute conversation with Rufus, Claude, or any other thing. If I already know what I want or I have a good idea, to type that and see the results instantly, nothing is going to beat that as far as having some two-way conversation. Where I think people are going to start, I'd like to get your opinion on this, but my personal opinion, none of us know anything. We can just make guesses and stuff. We don't know what's going to happen in the future. But my opinion is the thing that AI in Amazon or in general is going to change the shopping experience on is traditionally if you don't know what you want, the traditional path is, hey, let me go to Google or Bing or whatever and start doing research on it. Like my kids are doing a We're going to be talking about a theme party here next week. It's a – I think it's a tropical theme party. So they might do stuff like start Googling, hey, tropical theme party ideas or what is the best – what are the best decorations to have if I'm doing – like they're literally having kind of like conversations even in Google. Andrew Bell: I'm not even talking about using ChatGPT. That's a perfect example. Bradley Sutton: And I think that if Rufus and other things get more sophisticated, Google is now out of the question. You can now just go to Amazon and start in Rufus some of these conversations when you don't know what you want. You're like, hey, Rufus, what do you think I should get? I'm doing a tropical theme party. What do you think I should get? Oh, you should do this and that. That to me is what the future of search is. Not necessarily the shopping experience is going to change, but the initial research that you might have done is going to change. What are your thoughts? Andrew Bell: Now, I think that's an interesting point and I think like with this, it goes beyond Rufus. I want to touch on two different points. One that you said with you said the theme party you're doing. Okay, but one of the GBTs I created was actually based on a paper recently came out the Amazon science and it's backed up by most of the authors that come from Both the Rufus team, which there's, I think there's 60 people now on the Rufus team, which is much more than anything else. Like as an example, there's no one on the Amazon Inspire team. So it's like they got that out and there was only one person on Amazon Post team. So that was a telltale sign that they were going to go out too. But there's 60 people now on the Amazon team. In fact, they're just now hiring somebody that's going to be a manager for both Rufus and Alexa and how Alexa is going to take Rufus product intelligence. Because when you look at the job, you're seeing the future of what Amazon is creating. And so I really see what this science paper, it's really cool because it's taking that reasoning concept we talked about in product research, and it's actually going to apply it at scale on Amazon. Now, we're a little bit farther from that, but I've said that before and yet things accelerate at a pace that you couldn't believe. Reasoning models are expensive now, but they're going to be very cheap later on. Once upon a time, GBT-4 was really, really expensive and I could do 10 messages every three hours, but now I barely ever hit the limit on GBT-4. So all that to say with your tropical style theme party, you could type that into Rufus. It'll reason through. Everything that you've ever searched, right? Making assumptions of how you ask questions of how you do filters and bring that in into a cohesive whole to give you everything that you would need for a tropical style or at least suggestions. It's not saying everything that, oh, this is like something that we picked for you that we didn't think that you could do in yourself necessarily. But I think you'll be surprised by like having that proof of work with the reasoning models married to having products at every stage of question. Bradley Sutton: Okay, this is interesting. Like we can go on and on. This has already gone on too much because it's good information. But guys, we're going to have Andrew on maybe every other month here. Normally, we only have people on once a Once every year on the podcast and for most everybody we do, but those who are working on special projects with us like on a certain theme like AI is a huge thing for Helium 10 over the next year. So we're going to try and have Andrew on like maybe every other month on a live show or a podcast where he can come on and tell us what's changing because two months from now, Who knows, 20% of what we talked about today might be different because the world of AI, Rufus, and everything else is changing so rapidly. Andrew, we actually dropped your link in our YouTube version of the video, so we're going to drop the link so people can see it in the YouTube version to get these custom GPTs, but how can people find you on the interwebs out there just if they want to follow you and see what you do a little bit more? Andrew Bell: On the interwebs, LinkedIn. LinkedIn's the best. Just type in Andrew Bell. Hopefully, I show up. I hope I look like my picture. Bradley Sutton: Love it. Love it. Love it. All right. Well, Andrew, I would say Eucharisto for coming on the podcast today. I know I butchered the pronunciation. I can't use the excuse that I'm from Kentucky, but hey, I'm from California. We might be not as smart as those in Kentucky. But anyways, thank you for coming on the show this time and we look forward to, in a few weeks, bringing you back.