Amazon AI is Changing the Rules of the Search Game On You, with Spencer Millerberg, Founding Partner at DetailPage.com

Transcript

Our transcripts are generated by AI. Please excuse any typos and if you have any specific questions please email info@digitalshelfinstitute.org.

Peter Crosby (00:00):

Welcome to unpacking the Digital Shelf where we explore brand manufacturing in the digital age. Hey everyone. Peter Crosby here from the Digital Shelf Institute. Few weeks ago, Spencer Millerberg, founding partner at detailpage.com, was on our podcast talking about the latest in SEO trends in digital commerce. But we invited him back for a deeper dive on Amazon AI as they are the bellwether of where SEO will go in the next few years. The changes are ones no brand should sleep on, and the opportunities to win market share based on the right product page content to feed that hungry beast Rufus will be considerable. For example, Amazon has added hidden fields to fuel some of this new SEO and only 7% of brands have them filled out. Sure. SEO is just a game, but this time the rules are really changing. Spencer talked to Lauren Livak Gilbert and me about what's happening and what's ahead. Welcome back to the podcast, Spencer for part two of our conversation. Thanks so much for rejoining us.

Spencer Millerberg (01:13):

Hey, thanks for having me again. We're excited to be here.

Peter Crosby (01:16):

Just to refresh the poor listeners who somehow missed the first part of this episode, we started talking about SEO in retail, and now we're going to be digging into how AI is changing search, particularly really focusing on Amazon and where they're at is kind of the harbinger of things to come. I'm capturing it right. AI is, I mean, Amazon is really in the midst of making some changes happen and really recently at the Accelerate Show, and we're going to sort of dig into all of that. Is that the right focus framing for that?

Spencer Millerberg (01:58):

It is such an exciting time when it comes to Amazon. So many of the changes that we have been seeing them prepped for since even mid 2022 for two years plus, we're seeing them prep and make this roadmap and now some of these really needy changes are coming to fruition, is what I would say. So yes, you framed it up very well, and it's just a fun time to be living in right now,

Peter Crosby (02:26):

And I think it's the sort of rapid change that we're going to be seeing as AI continues to gather steam. And so I think all of that starting to sink in with folks is I think even part of the exercise that this can kick off. But before we go into those details, I understand from your LinkedIn, the congratulations are in order. You folks@detailpage.com just received a patent for AI assisted detail page content. Is that right?

Spencer Millerberg (03:00):

Yes, yes. We're super excited. So it's AI detail page content, and then we also have automatic keyword identification as well. So living in this new age of ai, you have to make sure that you have the right tools to capture the content that customers are searching for. And we're excited because it's a truly unique solution, at least the patent office thought so, and we're more excited because it's not just this cool invention, but we're getting insane results for our manufacturers who are using the service. And so we're very, very excited about the lifts in traffic, the 21% average lift in traffic that we're receiving when they have the right content. So thanks for noticing.

Peter Crosby (03:42):

Oh, no, and thank you for that stat because we're always looking for ways to understand what the impact of making the product detail page, because an investment, right, it's expensive in a lot of ways, and to hear something like 21% lift from improving your product detail page is pretty exciting.

Spencer Millerberg (04:06):

Patrick Miller and I, we were on the speaking circuit years ago at the same time, and we were both starting our companies and he had a great phrase that I loved. He said, you can always tell the brands that are going to be successful because of the ones who love their ASINs. And if they love their ASINs and they care for their ASINs, they care for their children, then that's what's going to make them successful. And that's exactly what we're seeing. It's that base content of loving your ASINs is what's happening here.

Peter Crosby (04:35):

I'm picturing a T-shirt right now, so we we'll work on or reality TV show. That's the one I'm to mind for me. Love your ass, love your As maybe next season. All right, then let's dig into the topic at hand, Amazon and ai. What is going on with how AI is influencing search on Amazon?

Spencer Millerberg (04:58):

Yeah, I mean, this is something they've been moving towards for a long, long time. Like I said, mid 2022, they started issuing this keyword by volume report, and then they had this disastrous gen AI description writer that everybody hated. So they kind of fumbled around for a little while trying to figure out what it was, but it didn't stop them. They're getting into some other things now that are better, like the AI review authentication. Is this a real review or a false review? Boy, that's been really beneficial to weed out some of those problems that have been there. They then started putting a bit more infrastructure behind things. They increased hidden terms, and then they added more fields. There's a key product features field that they launched in March of this year. They changed the indexing values and the way that the detailed page is indexed now, the title used to be completely separate and now it's all together.

(05:55):

So they have these backbones, and then you started seeing it from a consumer facing side with the Rufus search app that's in front. Not a ton of adoption there yet, but you watch Amazon is going to set the stage. You think about it, they are the largest e-commerce retailer. I mean, we start to see in a neighborhood of a hundred to one on the marketplace versus Walmart oftentimes. And so they really have this biggest and they're growing the fastest, and so it will set the customer stage for what they're going to expect, but I don't think they have it quite right yet. But they're getting a lot of really great progression,

Lauren Livak Gilbert (06:39):

And they added ads to Rufus too, which I think is a sign that they're trying to experiment with it. They want it to be an offering. They want people to start using it. They want brands to see it as an opportunity. I thought it would take longer for them to add ads, but it recently happened. So I think that's a big, not a red flag, just a flag that's waving that you should pay attention to. What do you think?

Spencer Millerberg (07:04):

Yeah, it's interesting because the AI is large language models and these things are super, super expensive to do, and so it makes sense that they're going to add it into their ads platform because if you think about it, this ads platform, it starts to become even a little bit more authentic where you're saying, Hey, what is the best thing for this situation? And the ad is serving you up, that's being served up to you is customized to the specific question that you're asking. It starts to become a little bit more authentic. I'm not quite sure they're going to get it right the first time, but I think that it will be more genuine versus hundreds of pages of search. And you think about that's the essence of what AI does is it's taking this massive, massive, massive amount of information and it's slimming it down into something that is much more, it's fewer, bigger and better. That's more specific to you. So I think that ads is going to stick, but I think they're going to have one or two more iterations before they get it quite right.

Peter Crosby (08:11):

Yeah, I definitely hear this reminder that this stuff will, I mean, that's part of the learning in the language model. It will get better over time, but that means that our listeners will be the Guinea pigs, as will the consumers as they go through this. And do you have a sense from what should our expectation be about how bumpy that ride might be as Amazon goes through this testing and learning process?

Spencer Millerberg (08:45):

Yeah, I don't know that it's going to be as bumpy for the consumer as it's going to be for the manufacturers and brands that are there, the manufacturers and brands, they are going to get a bit of a backlash where Amazon's only going to do things that are going to be beneficial for the consumer. I'll give you a couple of examples. When you think about Google search, I actually get kind of ticked off now when I search on Google and I have to go to the page and read, instead of having Google give me the ai, very few things right at the top, and you can see Amazon starting to move towards that. They haven't quite moved there with their roof of search, but it's going to change the way that the customer is able to get that. Now to the brands, that's going to be a tricky thing.

(09:43):

It's going to be a bumpy ride because if you don't have the right content, then you're not going to be one of those three or four options that the AI automated search is going to display for you there. And we're finding that written content is so much more indexable and searchable by the ai that becomes the predominant piece. And I would tell you that I told you that Amazon introduced a new field in March. We have 7% of manufacturers that come to us have that field filled out beforehand. And it's not like we're talking, our clients are the largest brands in the world that are using our services, and 7% of them are using these new fields, and so they're just not ready yet for this new AI revolution. So that's why I think customers, they're going to get the good side brands. I think they're going to struggle for a minute.

Lauren Livak Gilbert (10:37):

That was my question about Rufus. Do we actually know how it works? Because I know I've done a couple of tests myself, and when you ask, I asked something about skincare. I was like, I have sensitive skin. Can you recommend some lotions for me? I did the same sensitive skin lotion search and the search bar and Rufus didn't even give me the top 10 results. So do we have an understanding, and you said maybe it's written content of how Rufus is recommending because it doesn't seem to correlate with bestseller sponsored most frequently bought any of that.

Spencer Millerberg (11:10):

So Amazon hasn't published exactly what their large language model does, but we've done a lot of testing and we have a pretty good sense of what's happening here. What I would say is it's really, it's a very hungry, hungry hippo, right? Who's trying to eat all of this data and then trying to simplify the outputs that are there. So your query might have been something very specific about that sensitive skin, and it's likely going to go out. And we know that it crawls all of the written content on your PDPs. We know that it doesn't crawl a plus content, which is interesting, that is non-indexable field is what we're finding. And then there's some people who say that it falls the actual brand websites and Reddit forms, but although we're finding evidence to the contrary to that, that it's not necessarily doing that, and so it's grabbing this information. And so my guess from what you're saying here, Lauren, is that the algorithm giveth and the algorithm take it away. And you're probably seeing that the search algorithm is a bit more advanced in refining those exact things, and manufacturers have pushed on that. And so what we're seeing here is that this new large language model is grabbing content that is written in those people's pages almost specifically to be what it is that you are searching for. And so it's grabbing those fewer results and returning those fewer results to you.

Peter Crosby (12:43):

So besides Rufus, where else? Where is Amazon AI showing up in the shopping journey?

Spencer Millerberg (12:53):

Yeah, I mean the AI shopping journey, there's about four things that we're seeing that is really customer facing right now. There's a whole bunch of stuff on the backend, but you see four main pieces from the customer side. You see Alexa on their voice. We all know and love telling dad jokes and getting the top music. You have your Rufus search that we talked about. That's slim adoption so far. But I think that it'll start to move in mainline once they get it right. To your point number three, we're seeing this review summarizations down at the bottom where it's saying customers like this about the product and it gives you the top features, attributes of the product down below. And then number four, this is brand spanking new, just hot off the press last week at Amazon Accelerate. They stood up and they announced that they're going to be doing new dynamic titles, meaning that I search for something and I'm going to get a different title than Lauren's going to get, than Peter's going to get totally different. Very, very interesting.

Peter Crosby (14:00):

Wow. That's I think, exciting and frightening at the same time. What was the response like in the room?

Spencer Millerberg (14:10):

Oh my gosh, Peter, Peter, I'm telling you, there were audible gasps in the room when they announced this. It was insane. Everybody was like, oh, no, because the forums are so funny to read about this because it's saying, Hey, Amazon messed up my categorizations. Why not mess up my titles too? What's interesting, they gave an example on stage. They said, okay, you search for pink aviator sunglasses, and then they showed how they're going to change the title to put pink aviator sunglasses in the front of the title, and they're reordering those keywords. Okay, here's the interesting thing. Aviators a trademark term by RayBan. So they violated on stage in their example. Oh boy. And it showed that they're not quite getting it right in their prime example, that they're showing us their best case, their best use case, right? So there's going to be some bumps in the road is what I would say. But I think that manufacturers and brands cannot change the rules of the game. They have to get prepared for these new rules is what I would say.

Lauren Livak Gilbert (15:26):

And how do they get prepared for them? I mean, the gasps say it all brands are not ready for customized titles. Some of them aren't even ready for the titles that they have to put on just to kind of meet the needs of Amazon to set up an item. So

Spencer Millerberg (15:42):

What

Lauren Livak Gilbert (15:42):

Should they be doing? What should they be asking? How should they be preparing?

Spencer Millerberg (15:46):

Yeah. No, that's a good question. So Amazon is very specific in the things that they say, right, order in the next three hours and seven minutes and you're going to get it tomorrow. So they're exact and precise in their wording. And if you look at what they put on the screen and exactly how they said it, they said that they are going to intelligently reorder how words are displayed. So it gathers customer preferences, and he used the example of search and then it reorders how words are displayed. So really what do brands need to do? They need to know what do customer search for number one. And then they need to make sure that the words that customers search for are in their content. Because Amazon cannot create new content based upon what they're saying here. They're not going to just, if it doesn't say aviator sunglasses, it's not going to create aviator sunglasses. They're going to reorder that. So if you search for a photo printer and you have that in your content, then they can put that in the title and it will elevate it within search, but it won't add photo printer to your printer if it doesn't have photo printer and include it in it already. Does that make sense?

Lauren Livak Gilbert (17:01):

And correct me if I'm wrong, but it won't change when they click to go to the PDP, it will only change in the search results. So they're not changing your PD DP itself, they're just changing the initial view in the search results. Is that correct? I could be.

Spencer Millerberg (17:17):

They have not defined that yet. They haven't said if it's going to be changed when you click through to the PDP or not, but they did say that it changes within the search results. The example that they gave had aviator at the very, very end of a title, and then they pushed it to the front. But what we're seeing about three, four months ago, Amazon made a major change to their search function. It used to be that the title was indexed separately than the bullet points were. And now we've seen that their title and their bullet points blend together. They can canton them together, is a functional term. And so that's interesting because for a branded manufacturer that has a couple of implications. Number one, you don't have to keyword stuff your title anymore. Number two, your first bullet points that you have, they have to make sure that you're using that customer search language, not flowery generic or conversion based language that we have here. I see all the time people say things like works in all washing machines. Well, guess what? No customer ever searched for works in all washing machines.

Lauren Livak Gilbert (18:25):

Most customers don't even know the brand of washing machine that they have.

Spencer Millerberg (18:28):

Exactly right or recommended by dermatologists. Guess what? Again, nobody searches recommended by dermatologist when they're talking about their skincare or their laundry detergent or anything else. So you have to make sure you have the right search terms up higher within those bullet points, and that's going to gather that preferences for you. Because I think what's going to happen, Lauren, is they are going to maintain that title when you click through to the detail page, and they're going to pull content now from those initial bullet points in order to reorder and put within the content as well of the title. Because the titles as they stand right now, they're too slim. They're not Chinese importers and manufacturers are killing it, but most of the branded manufacturers, they don't have the content necessary to do this.

Peter Crosby (19:19):

So I'm guessing because Amazon has their ai part of the answer to do this certainly at scale has something to do with your patent that we talked about the opening of the show that if we can't go back to the days where you have a human being trying to do this on just your top 10 pages, that's not going to fill the bill. And so I'm promoting your product here just because this is a new problem and a new opportunity that needs to be solved that really just can't be solved by throwing people at the problem for brands that are now trying to work profitably at this process. Is that fair to say?

Spencer Millerberg (20:06):

Yeah. Well, hey, thanks for the plug. Appreciate it. No, Peter, I do this because I want to do this, and I found that within the market, this was a huge gap. When I was waiting out my non-compete, I went to work for Blackstone, and so I was running a whole bunch of data with my partner, Justin Mayr, best guy in the world. And we found that when you optimized the base level content, you could compete really effectively within Amazon. And so we saw that even before this AI search came into play, and now it's even more important. And then what we're seeing is that there's a change that happens on a quarterly that you have to change your content about quarterly because of three reasons. Amazon's changing their search algorithm on a constant ongoing basis. They're adding those new fields. Number two, you have the seasonality that exists even in innocuous keywords like toilet paper you're getting, people will say, Hey, bulk toilet paper or family sized toilet paper, they have keyword fatigue and it changes over the time.

(21:13):

But third and probably most important, what we're seeing is that Amazon has a little bit of, they have this honeymoon period when you update your content. They want new and improved products to float to the top. So they give you this about 90 day benefit for having new and improved content, but you have to update it. At least 13% was the threshold that we found of content. So yeah, you just can't get that frequency of update and you can't get that the right keywords without having either a really, really cheap offshore labor, which most of the time can't do the mathematics or number two really good software solutions that are there. So yeah, the short answer is yes, it's absolutely something that's needed.

Lauren Livak Gilbert (21:57):

So if enough hasn't already changed with all these announcements, where do you think AI is going to continue to take us from a search perspective? I mean, I fundamentally believe that's the biggest place that it's going to impact how we search. So what are you seeing moving forward? What's catching your eye? What should brands be looking

Spencer Millerberg (22:14):

For? Yeah, so we talked a little bit about advertising and I think that advertising will be that next play. I think that they're going to see within advertising, personalization is going to become the next play within that. It's going to understand better what you purchased before, all the widgets that Amazon already has of other customers who bought this bought also bought. That's going to go to the next level. And then I also think that you're going to see that starting to move within the search results on the organic side, that search results are going to go in line similar to what Google has, that you're going to have those AI recommendations from Rufus no longer up in an option to search. It will become the main search and it will become part of those widgets that we have of Amazon recommended, et cetera. That's where we're going to change. And you're going to start to see, oh, you bought an HT printer. Here's the ink that fits that. Oh, you bought dog food clips, clothes, et cetera. This is the personalized things that fit better with you, Lauren, as opposed to just the general search and the general public.

Lauren Livak Gilbert (23:25):

And do you think it'll go more towards, I'm calling it mission-based search. I've heard a lot of people call it different things. So for example, I can't believe Halloween is coming up, but it is. So for Halloween, it's like, I need to get my kid prepared for Halloween, and it'll pull costumes and candy and a bag for the candy and multiple different categories together. Is that where you think it will move towards?

Spencer Millerberg (23:47):

I don't think that, no, because when you go to the store, you don't necessarily shop that way. You kind of know what you want and you know what you're looking for. When you think about your Halloween example, my family is dressing up as a Harry Potter characters. I have a 9-year-old and she's obsessed. I love it. And so I know exactly what I'm looking for when I'm coming in and my search is much more specific now. I think that the recommendations around that will be tighter in and AI enhanced because of the background. They're going to know my sizes. They're going to know what specifically I'm looking for there. And so those will be tighter, but I don't think it'll be more of that broad occasion based. I just don't think we're going to move there because the customer doesn't move there. That's not how they search.

(24:37):

Yeah. Now, I'll give one caveat to that though. We are seeing occasion based searches become really important when it comes to gifting occasions especially. So if I were DeWalt and I have my power drills, gifts for dad would be critical types of keywords that you have there. We see in toys, some of the most important and most essential keywords on toys are age specific and gender specific. So 4-year-old birthday gift, that sort of thing is very, very high volume in search that you have customized M and MSM and pink m and ms gender reveal party treats and things like that. Wet. Oh yeah, ding.

Lauren Livak Gilbert (25:28):

Awesome. Oreos. That was my favorite.

Spencer Millerberg (25:30):

See, even in medications, you think that makes sense, but then in medications, we're seeing things like there was a dry mouth lozenge, CPAP supplies and those for people going through cancer and radiation and chemotherapy, dry mouth, those types of keywords really start to become more critical to have that occasion based information. In your detail pages,

Peter Crosby (25:58):

What do you think the timeline is for this personalization of Amazon's widgets? Do you have a sense of that in your crystal ball?

Spencer Millerberg (26:06):

Yeah, my crystal ball. I mean, let me check it.

Lauren Livak Gilbert (26:09):

Or is it a magic eight ball

Spencer Millerberg (26:14):

Begin later is what it says? Exactly. No, I mean, look, here's what I'd say is this is going to happen faster than we think. I think it's going to take two to three years to get more into that next steady state. I think we're going to see a huge amount of trial over those in between now and then, but I think it's going to take about three years to get there and get really, really tight within there. Maybe on the outside of chance five years, but I think it'll be closer into three.

Peter Crosby (26:47):

Well, I mean, in some ways, I mean, I know all of this is happening right now, and I'm sure our listeners feel a little bit of a clench in their chest, like, oh my God, what's happening? But that says to me that there is a timeframe to work on this problem, that it's not just do something about it now or you're screwed. Yeah. And so I think it'd be great to close out just with what should they be doing starting right now or after breakfast to do to get prepared for that future? What are the ideas that they should have in mind as the best practices here?

Spencer Millerberg (27:32):

Yeah, absolutely. Look, this is just a game, but the rules of the game have changed under our feet now, and big brands have historically been slower to adapt than their offshore counterparts. And so as they adapt, I'd say move quickly. And the things that you need to do are, number one, it's use customer search language. If you don't use the customer search language, the crawl won't work. The adaptive titles won't work. You will end up getting beaten by those who are moving towards that customer search language in this new AI enhanced search world. Number two, you got to get specific for those use cases. We gave the example of toys for 4-year-old boys individually wrapped snacks for lunches fan for college dorms. You can't include content if it's not there, but it does become much more able to recommend a fan for your college dorm if you say it right in your content that you have there.

(28:30):

Number three, that's timely updates. So that fan for college dorms might need to be adapting during the winter time when it's not a college dorm purchase season, and you might have different terms that you have there. What we're also finding is you have to update your content approximately six weeks before major events, like a prime day or a holiday period of Black Friday, because as you update it, these large language models, they take a little bit of time to learn and adapt, and you need about that six weeks in order to get there. And then last I would say is use Amazon's new reporting. There's tons of great new reporting out there, and most manufacturers are not using it. They're not adapting to fit what's working and then adapting and kicking out what's not.

Peter Crosby (29:19):

Can you double click on that reporting item for me just a little bit if it's in your head? Just what are the ones you think customers really ought to be paying attention to?

Spencer Millerberg (29:31):

Yeah. Yeah. So Amazon introduced a keyword volume by term report that is included there, and then they just introduced some more, oh, I can't remember the name of it. It's in the international section that they have where it's talking about volume over time that they have in there. So there's a lot of ways that you can triangulate the data. Amazon never gives you all the data in one spot. I mean, why would they, God forbid, makes it too easy, too easy. But if you use three different reports, you can start to triangulate exactly what it is that you need.

Peter Crosby (30:09):

I always keep hoping that because Amazon has people in the private label business who have to sell on Amazon, that there might be inside voices who are like, guys, come on. You got to put this information in the same spot, and you've got to, because I don't know that they always understand what their brands are going through to try to keep up with all of this. But you never know. Maybe an inside voice will help. Well, Spencer, if only,

(30:39):

Yes, if only Spencer, grateful as always for you to do a twofer with us in the past few weeks. We really appreciate it, and particularly on this topic, raising the early and pretty important alarm of the way the rules are changing and then the opportunities they're going to come out of that for brands that figure out the way forward in a way that they can afford to do and in a way that they can show measurable results for that work, I think is pretty exciting. So thank you so much for coming back and sharing this with us.

Spencer Millerberg (31:14):

Hey, thank you for having me. Amazon's reset the board again, and now we just have to play their game, so it's going to be great.

Peter Crosby (31:21):

Thanks again to Spencer for the valuable heads up. The next few years won't be dull. Don't you want to be ahead of it all at the DSI? Go to digitalshelfinstitute.org and become a member for free. Thanks for being part of our community.