← Behind the Numbers: an EMARKETER Podcast
Behind the Numbers: an EMARKETER Podcast
Amazon’s AI Overviews Changes Discovery? Shopping AI or General AI? 3 Big Questions for Amazon | Behind the Numbers
00:29:03
Loading summary
Transcript104 lines
- [00:00]
A
Foreign.
- [00:05]
B
Hey, gang. It's Friday, May 15th. Marissa, Susie, and listeners, welcome to behind the Numbers New Marks podcast. I'm Marcus, and joining me for today's conversation, we have two New York people. At least they live there. Newsletter analyst and writer Marissa Jones joins the show.
- [00:22]
A
Hi.
- [00:23]
C
Thank you for having me.
- [00:24]
B
Why, of course. Of course. We're so joined by VP of content and host of our reimagining retail show, Suzy David Kenyon.
- [00:31]
A
Thanks for having me.
- [00:33]
B
Welcome, welcome. We start, of course, with the fact of the day. Okay, what's wrong?
- [00:48]
A
I'm just. My palms are sweaty. I'm waiting to see. Plus, I would like my window to be open, but I know there's too much noise in New York.
- [00:56]
B
All right, here's my fact. Where did the phrase mad as a hatter.
- [01:01]
A
Mad as a what?
- [01:05]
B
Mad as a hatter. Have you heard this before? You've not heard of this before? Marissa has. Have you heard of this?
- [01:13]
A
I have not. But in Canada, which I don't know if this is where you're going, we used to have a birthday party place called the Mad Hatter, and it was like a kids sort of like fun land.
- [01:23]
B
Well, so there's the character from Alice in Wonderlands called the Man Hatter, but there's the phrase mad as a hatter. Oh, they're mad as a hatter. And so it originated in the. In the 18th and 19th centuries due to mercury. Due to mercury poisoning among hat makers, and they were called hatters. They used mercury nitrate to turn fur into felt, and chronic exposure caused tremors, speech issues, and hallucinations, leading to irrational behavior, often interpreted as insanity. That's where it came from.
- [02:02]
A
Is this where I ask, what does that have to do with me?
- [02:05]
B
What do you. What are you talking about?
- [02:07]
A
Leading in. You were like, this one's for you.
- [02:10]
B
And I never said that. I said. I might have said that. I said. I think I said, you're going to like this one because it's interesting. Isn't that what I said?
- [02:19]
A
Yeah.
- [02:20]
B
Okay, that's probably what it was.
- [02:22]
A
I thought you were going with sushi and mercury poisoning.
- [02:25]
B
Sushi?
- [02:26]
C
Yeah.
- [02:27]
A
Too much tuna.
- [02:28]
B
Oh. Oh, no. Hat making. This happened long before Lewis Carroll popularized the character. The Mad Hatter in Alice's Adventures in Wonderland was published in 1865.
- [02:42]
A
Oh, wow.
- [02:43]
B
Whoa. Anyway, that's where the term matt is a hatter comes from.
- [02:48]
A
Marissa, had you actually heard that expression before?
- [02:51]
C
Yeah, I had, but I thought it was just from Alice in Wonderland, so I learned something new today, but I have heard it.
- [02:57]
A
Yeah. I'm, I'm going with. I'm from Canada. You didn't use that.
- [03:03]
B
Today's real topic, the big three questions surrounding Amazon. All right, we set the table. First, how have they been doing? Within Q1 2026, Amazon made 181, $80 billion in total revenue. That's a lot. It's up 17% year over year. It's double last Q1's growth. Online net, store sales, that's its largest slice, but it's down to 35% share of the total. But it did grow down in terms of the share shrunk, but it's still growing. As a line item, it grew 12% also double last Q1. AWS is now a 20% slice of the total revenue that grew 28% in the quarter, the fastest in nearly four years. And ad revenue was close to 10% of its earnings. 10% slice that grew 24%, the fastest quarterly growth in two years. But we're here to talk about three big questions surrounding Amazon. Marissa, what is top of mind for you?
- [04:06]
C
Okay, so my first big question is we're seeing Amazon's ad revenues are growing consistently at a double digit pace according 1 to their earnings, but also to our forecasts that track their yearly ad revenues. So we're seeing them growing at a double pace, but the overall volume of these ad revenues is still very much notably behind Amazon's two biggest triathlete competitors, Google and Meta. So my question is what would it really take for Amazon to actually close that gap or come much closer to closing that gap?
- [04:40]
B
To close in the gap of, of
- [04:43]
C
between Meta and Google in ad revenue volume.
- [04:47]
B
Oh, okay. In terms of like they're part of the triopoly, but could they start to threaten Google? Yes, which is now I guess in second place. But I mean it's close. Well, in terms of nets at least. Yeah, growing pretty healthy. Double digits, right? 80% this year for ad revenues in the US 17 next year, dropping to a close to a 14 a year after that. What do they have to do? Well, why is this hard took? Go see, please.
- [05:20]
A
No, no, I was just going to say when you're already quite large and you're a little bit behind potentially it's really hard. But it sounds like if they continue to grow their retail business, which is that like centerpiece that helps prime, and if Amazon is thinking about all of the different off site mechanisms and all the different partners, it's probably just a matter of time that they narrow the gap.
- [05:45]
B
Marissa, why this question? Why now?
- [05:48]
C
I think this question, we've kind of been tracking this idea that Amazon is often growing at a faster rate than both of them, I believe only faster than Google. This quarter I think Meta outpaced their growth. But we very much position like the big three tech companies, the big three ad earners as these three. But there is still even despite their growth, ads are only really a fraction of Amazon's overall business. But they are simultaneously making a very heavy ad push. They're coming out with AI tools to do everything for you. They're really pushing prime video ads, they're introducing ads in their AI chatbot Rufus and pushing that as a high intent environment. So you can really see that they are also trying to kind of legitimize their position as not just a retail media platform but as a sustainable source of ad growth.
- [06:42]
A
Well, I think their partnerships with folks like Macy's too will probably help them grow and become even more, as you're saying, legitimate as an ad platform beyond their own space to help facilitate other people's buys
- [06:58]
B
currently. So we have Meta 25% digital ad revenue in the US Google 24% Amazon 14 but still growing. Google is kind of shrinking a bit. Meta is ticking up but yeah, in a couple of years Amazon capturing 15% and that'll be compared to Google's 22. So it's getting there.
- [07:26]
A
Yeah, absolutely.
- [07:27]
B
Yeah, it's a good one. Susie, what'd you have for us?
- [07:32]
A
Can Amazon become the operating system for retail as it relates to supply chain? So think about the awsification but on supply side like couple of weeks ago, trying to make the dates happen in my head. A couple weeks ago Amazon announced their newest Amazon supply chain services which for those of you who did not hear the news, it is about end to end fulfillment from not just the last mile and it's beyond their current clients or you know, players or sellers. It's like way beyond the sourced PNG3M as examples of who they might work with. They're adding a data AI layer and it's really meant to be way beyond. They already had fulfilled by Amazon. They have buy with Prime. They have lots of different components of the fulfillment process but this is really to get to the root of supply chain. So you know they've learned from their playbook, build something inside house aws, figure out how to monetize it and then start selling it to others. I think this one's a little harder because it's so capital intensive versus aws. So will they be able to make money on it or is it just a way for them to control more of retail in different ways.
- [08:56]
B
What do you think? Do you have a take on that?
- [08:59]
A
I mean I think it's very clever because the more pots of supply chain that they're in, the better it is for them. There are probably learnings that they're making and they have different relationships that they're building with the retailers and brands. I don't know if it's sustainable, but I'm not surprised that they're trying. They're very good at do it yourself, roll it out. Oh, didn't work. Let's pivot until they find the right formula or just move on with their life anyways. They've already spent the money trying to figure it out for them. Helps.
- [09:32]
B
Yeah, very nice. Marissa, back to you.
- [09:37]
A
What?
- [09:37]
C
Do you have another question? Okay, so my next question is kind going back to Rufus. So could Rufus turn AI Shopping assistance into one of Amazon's most valuable high intent ad services?
- [09:51]
B
Oh, okay. This is close to one of the ones I have. Keep going, tell me more.
- [09:57]
C
So obviously we're seeing Amazon really kind of push ads into Rufus. There was recently a pitch deck that leaked, I think that Adweek reported on essentially showing Amazon making the case for Rufus ads to its group of advertisers. They're really rolling them out and they're positioning this as a point where users are actively looking for a specific product or a specific solution to a problem that they're having. Rufus can service these and it's a very high intent environment. Amazon saying it converts shoppers. I'm sorry, it converts users into buyers pretty effectively. But how much of that is of Amazon's future? Ad business is going to be driven by Rufus and its AI powered chatbot versus what else it's doing in that.
- [10:46]
A
It's funny, me too. I thought about this one but luckily I have a different question. I think that it's going to replace their on site business, ad business, but not their off site. And they might have even more data to strengthen the entire ad business.
- [11:05]
C
Yeah.
- [11:08]
B
So this pivot into one of the questions I had which was will customers? Because to build an ad business on it, you need people to use it, Right. And they are. It seems Rufus has seen it. Was saying Rachel was writing a piece. Rachel Wolf, one of our retail analysts was noting that Rufus has seen rapid adoption. Amazon noting the tool was used by 250 million folks in 2025 and pointing out that users interacting with Rufus was 60% more likely to complete a purchase. So it is being used. It Is it does seem like it's effective, even though some kind of very early kind of trial stages, it seems. My question was, will customers want a shopping specific AI to help them like a Rufus or a Sparky or a generalist model like a chat GPT or Gemini? So is there any thoughts there on whether because they're building this thing, it kind of lives inside the Amazon world. That's a lot of the time where people go to shop. But could you see people's shopping behavior kind of being pulled out of that environment and then wanting to use the same generalist tool that they use to copy edit an essay for them or to ask them a question about something going on, something to do with history, to use that same model to also buy them certain things.
- [12:30]
A
So I mean that was a lot of questions. In one, I think Rufus is being pushed into people's systems. So like when you're on the app, it's like Rufus sort of kind of sometimes just pops up. So whether people are choosing to go there or not, they're definitely being sort of led to Rufus in many instances. And it's even for the returns process, it's for. For the entire SH shopping journey. So I think there, it's not a surprise that there's such an uptick. The idea around do I want an LLM or do I want Rufus to purchase something for me is a little bit different from an assistant. Right. The assistant is just helping me find the answers that I want and then I make the ultimate decision versus do I want an LLM to purchase something? So like truly autonomous agentic commerce. I don't think anybody is there yet whether it's Rufus or not Rufus. The closest Amazon has to that is the subscription business, which is not actually. Well, it might be AI driven on the back end, but it doesn't feel like AI on the front end where you're like every three weeks I want you to send me whatever it is that you need for your household and it gives you the little coupon for subscribe and save. So there are some instances of I relinquish power and you buy for me. But I don't think that's what you're asking. You're asking, would someone prefer to be on Rufus or on an LLM to decide what they're going to buy and have their own decision making power. And I think yes.
- [13:50]
B
Would they say to themselves, okay, pause, I'm gonna go, I need to shop. I need to go to the shopping specific AI for that?
- [13:57]
A
Yes. So I think when you are an Amazon prime or a Walmart plus member and you are paying for your loyalty. You're paying them to be loyal to them. The chances that you go to Sparky or Amazon to repeat purchases is pretty strong. But when you're looking for something brand new, or you're looking to be inspired, or you're thinking about like it's my birthday and I want a theme, you might go to your favorite LLM and do it because then it'll be like, okay, you need party favors from Party City, which is at Staples, and you need whatever dips from this company, but then you need that from that other company. So then I think an LLM will be easier. I don't know that there's a world where people are going to choose one or the other. I think they're going to do whatever makes their life easiest.
- [14:44]
B
Yeah. Um, Marissa, what do you think? The. The. The Economist had an interesting line saying one source of worry for the company is is the rise of AI agents embedded in chatbots that can shop on a user's behalf. They could threaten Amazon's core business, bypassing Amazon.com or robbing it of relationships with customers and advertisers. Yet the company is better placed than its skeptics fear. Shortly after it struck a deal with OpenAI, the model maker shelved plans to launch a shopping service called Instant Checkout. Rufus has also helped generate $12 billion in incremental annualized sales last year. So what were your thoughts here?
- [15:22]
C
I think there is some interest in a shopping specific AI tool like Rufus, mostly because it kind of feels like a normal chatbot. Before Genai was really a thing, it's just a virtual assistant. I think it doesn't really feel as maybe intimidating to users who are kind of averse to AI. I think Amazon has a good benefit that everything could happen within its own ecosystem versus a platform like ChatGPT where it can give you the recommendations, kind of lead you to the checkout, but not necessarily complete everything yet within its own ecosystem though there are probably moving in that direction. But I think for now Amazon has quite an advantage and I don't think users would be as averse to it as they would with a different LLM.
- [16:09]
B
We do have numbers on generative AI shopping users, so this is using either the Chat GPT or Rufus or a Sparky or something, something like that. We have about 80, 80 million in 2026. That is about half of Gen AI users and about a quarter of people in America. And that's this year. It's growing probably about 8 to 10 million over the next couple of each of the next couple of years. Probably 6 to 8 million each next couple of years. I have another one for you guys which I want to throw out there because I think it's related. And this one came from Rachel, Rachel Wolf, who I mentioned earlier, our retail analyst. She wrote a really interesting piece and the question that came out of it is, can Amazon's test of AI generated search overviews reshape discovery? So AI overviews, you know them as the ones that Google has. When you type in something, it gives you a little summary box at the top and these apparently are coming or this style of search responses coming to Amazon. She's writing that Amazon's AI experiments could radically transform the retail search experience. Instead of being confronted with a large array of similar looking products, shoppers may be shown a smaller selection of items, three products along with backup options better suited to their needs. Suzy, do you think Amazon's test of AI generated search overviews could reshape discovery?
- [17:33]
A
So I think, I don't mean to hedge, but I just think it depends on the category. If it's like an everyday household item that you're constantly buying, you probably will even skip overviews and you'll just go to whatever you're buying to make your life easier. If you're kind of open and on the fence about the brand, you don't love the last experience you had with that brand. I think it'll help a little bit, take away like change discovery, but I don't think when it's something net new for you or it's something that's a trend piece or like clothes where you're, you know, you have a specific distinct style. Certainly not in the beginning when it doesn't understand your style because it's kind of like furniture shopping, right? You can ask for a blue couch and you can get all kinds of different things that show up. So it's like they have to refine it and I don't think an overview is going to be able to do that in one shot, but it is definitely. Everybody, every retailer needs to take a step back and think about the impact of AI to what they're doing and whether the tools are customer facing or not to just make sure that they are part of the change.
- [18:34]
B
Yeah, yeah. Marissa. Rachel illustrates some of the pros and cons of this of the AI generated search overviews for Amazon. On the one hand, she was saying it could deliver more curated results, helping to minimize decision paralysis and consequently increase conversions on the other, showing fewer products means fewer advertising opportunities, which could be a potential sticking point since nearly 70% of Amazon's US search ad revenues this year will come from search. Which is what Any thoughts on these search overviews for Amazon?
- [19:05]
C
It kind of sparks another question that I was kind of, that I was considering touching on, which is could Rufus and this new sponsors search push, I guess diminish the power of traditional sponsored search that advertisers have relied on for so long with Amazon? Does that lose relevance in a heavily AI driven environment?
- [19:32]
B
Yeah. Susie, any thoughts?
- [19:35]
A
Well, I think these AI overviews like are a different way of saying curation. Right. And like you need to show me as a shopper, you need to show me what's relevant to me in the moment that I'm ready to make the purchase or else I'm going to walk away. There is a lot of, and we talked about this with Rocked, like there's a lot of endless aisle. The bigger a marketplace is, the more stuff is in the pages for you to try and find. The job of the folks that are programming the AI overviews or any sort of personalization quote unquote hack is to make sure that they give you enough friction so you discover new things, but not so much friction that you leave. And so I think it's like that midpoint of the overview will make my life easier so I don't have decision paralysis, but also give me a few more options that maybe I didn't have in mind. Yeah, and I think that's just going to take some time. The LLMs are learning about each of us as they get more information and they're learning more about the retailers and the brands and so all of that kind of coming together will make a difference.
- [20:39]
B
There's also a trust piece here because it's not just a result, it's a result that people, a lot of people are becoming quite skeptical of because these things can hallucinate and people are nervous that they're. The results they're getting aren't just organic results, natural results. They're results that are being paid for by certain companies, certain brands, certain advertisers. This chart folks can see on the screen some biz rate insights that I found. And it noted AI generated summaries or overviews of some of the least trusted sources for online shoppers. So that's customer ratings and reviews with the highest most trusted 71% recommendations from friends, family in second place, customer submitted photos or videos in third. And then you keep going down and with just 15% was AI generated summaries or overviews right there with influencer reviews. That could be because these are outside. Like if you had overviews that were out inside of Amazon, maybe they'd be a bit more trusted than the ones that they're seeing outside of Amazon because they trust that shopping environment and that brand, that experience. But for now, at least, some of
- [21:49]
A
the least trusted don't equate overviews with advertising. And I feel like they have to tell you it's kind of like the sponsored ads on Amazon. When you're searching, it's very clear that it's paid for advertising. What's annoying in the current search is that sometimes there's way more sponsored ads things before you get to the organic stuff. I wonder if people are mistrustful over the overviews when it comes to shopping because they had something specific in mind and that didn't show up. Or like they asked for Nike runners or they asked for running shoes and expected to see Nike and that didn't show up organic or not organic. And then they're like, oh, if the biggest brand isn't showing up, then I don't trust it. But I just think it's the way that the product pages are on the back end, that's what's showing up.
- [22:36]
B
It's a great point. And this is kind of. I always find this conversation quite interesting, is that we notice it, but I do wonder how much people outside of our world notice it. And so it's kind of like the right, this is it meta that right now they label their post sponsored. They're going to be switching that to just ad. And there's questions about will people notice it even less. But then there are questions about how much do people notice sponsored right now? And we know, because we know how advertising works, but how much does the average person know that this thing is sponsored? This thing isn't. This was influenced by an advertiser, this one isn't. But you're right, there are labels. So yeah, maybe, maybe that is already clear. Do you have Susie, you have another one?
- [23:21]
A
I have one more. Of course it's retail related.
- [23:24]
B
Sure.
- [23:25]
A
Is Amazon even a retailer anymore? In that it started as a bookstore. We know that it turned into a marketplace, the marketplace turned into prime. I might not get it in the exact order, but you know what I'm saying? Like, and then it became aws, and then they started selling, selling their fulfillment services. And then they started selling definitely not in the right order. But then they started selling space on their marketplace. Originally it was their products and then they had this 3Pmodel. Then they started doing advertising and then they're making AI chips. Like it's just ongoing. The list of things that they're doing that are no longer retail even. You know, they're like retail adjacent. So are they a retailer?
- [24:08]
B
It's a good one. And I struggle with this now when I'm talking about Amazon or like working on a script for them and every time I go to say the retailer, I stop myself and say the company or the anything and their earnings would agree with you. Look at, I mean quarter on quarter things fluctuates is now and stuff. If you look at their full year earnings from last year, like I was saying before, online net store sales is 38% now. And that was, you know, just going back in time. That was 50% in 2020. So that's already shrunk quite a lot. Then they're getting 24% of theirs from, from third party, third party seller services but they're getting 18% from AWS, they're getting 10% from advertising, they're getting 7% from subscriptions. So it's a good question.
- [24:59]
A
I mean for me retail is the underpinning, right? That's what's fueling everything and it's the engine that they need have to do everything else. It's very norm. Especially 2020 taught us they were already ahead of the curve. But like you need to as a retailer you need to have other revenue streams if you are only relying especially on stores or if you think about it like distribution centers were closed, like people weren't anywhere near physical spaces to even deliver your goods to you. So clearly it's important to have more than one revenue stream. But it's still, I mean it's still a big chunk of their. When you add 3p 1p. Because I don't again, consumers don't understand the difference. For them it's just. And when we think about their business, we compare the entire marketplace regardless of who's selling to a Target or a Walmart. So I think yeah, they're kind of a retailer but with lots of other sort of tentacles. I don't know if that's the right word.
- [25:54]
B
Irons in the fire maybe. Could be anything.
- [25:57]
A
Mad matters
- [26:00]
B
definitely. All right folks, let's pick a top three. Marissa, which of the ones that we've talked about do you want to keep? For our top three, I'm going to
- [26:09]
C
keep one of mine which is could Rufus turn AI shopping assistants into one of Amazon's most valuable high intent ad services.
- [26:16]
B
Suzy, how about for you?
- [26:18]
A
So I definitely think we need a Rufus one. I feel like there has to be an AI big question. I think if I was is that
- [26:27]
B
your choice or what's happening right now?
- [26:29]
A
I am playing within the borders of the rules.
- [26:33]
B
Okay.
- [26:34]
A
I'm agreeing that that should top question.
- [26:36]
B
Oh, okay.
- [26:37]
A
And I'm adding that. Yeah, here we go. And I'm adding that if I pick the top question based on relevance it would be the supply chain. Like are they going to be able to be the backbone of retail supply chains based on their newest announcement. But if I'm talking about the best question from a philosophical most chatter it would be are they still a retailer? So I don't know. I'm not sure. It's too hard.
- [27:03]
B
Five minutes later, I've got nothing.
- [27:05]
A
The supply chain one. How about that?
- [27:07]
B
Very good. Can Amazon be the backbone of the retail supply chain?
- [27:10]
A
Yeah, I mean that wasn't the original question but I think that's fair.
- [27:13]
B
Yeah, we do already have a Rufus one, so I'm going to skip over that and go to the one that I stole basically from Rachel because she's smarter than me. So I'm going to include this one. Can Amazon's test of AI generated search overviews reshape discovery? So our top three for today, big questions, big three questions for Amazon are could Rufus turn AI shopping assistants into one of Amazon's high. Sorry, Amazon's most high intent ad services. Then we had can Amazon be the backbone of the retail supply chain? And then lastly, can Amazon's test of AI generated search overviews reshape discovery? There are top three for you today. Thank you so much to my guests for helping me put them together. Thank you.
- [27:51]
C
First Marissa, thank you for having me.
- [27:53]
B
And of course to Susie.
- [27:54]
A
Thanks so much.
- [27:55]
B
Yes, indeed. And to Lance on the crew today helping us with this episode and to everyone for listening in to behind the Numbers knee marketer podcast. Keep listening to the end of these episodes to hear new outtakes and behind the scenes clips these guys are just learning about. We'll be back on Monday. Until then, happiest of weekends.
- [28:12]
A
It recorded when you said that I was amazing.
- [28:15]
B
Never said you were amazing.
- [28:17]
A
You did now.
- [28:20]
B
I don't know what it.
- [28:22]
A
Lance is gonna cut it. So it sounds like you said you're amazing.
- [28:29]
B
I do have a fact of the day which is I don't want to say just for you.
- [28:34]
A
Oh God. Is it about Canada?
- [28:36]
B
Should have said that. It's super offensive. No, it's not about Canada.
- [28:41]
A
Oh, then I'm offended. I'm already offended.
- [28:44]
B
You should be, actually.
- [28:45]
A
Oh, great. Perfect. I love it.
- [28:50]
B
Here we go.