Retail Pulse Report: Why GenAI Search Could Kill Brands

Google's new AI Mode isn't just changing search - it's making traditional SEO obsolete and putting brand visibility at risk

Source: Adobe Stock, “SEO obsolete”: as dead as you’ll be when Gen Alpha asks you what this is a picture of.

First, thanks for your patience as I climb back on top of the pile after an inspiring and frustrating trip to Honduras with RetailROI. More on that at the bottom.

Before we get there, though, rather than focus on the news (because I didn’t), I’m going to dive into a very hot topic: what GenAI is doing to search. Every once in a while you come across a piece of research and commentary that you have to read twice, and I found that in this piece by Mike King at iPullRank. If you want to understand why Google said it was a red alert moment when OpenAI launched ChatGPT, what they’re doing in response – through AI Mode – will explain exactly why.

However, what Google is doing in response is a red alert moment for the retail industry too. Let’s take a look at why.

It’s Already Here: Google’s AI Mode and the Death of SEO

Monetizing consumer attention is hard, and it’s extra difficult when the way that monetization is achieved is by spurring a customer to buy something that they arguably weren’t already planning on buying. A retailer or brand pays Google (and SEO consultants) to rank in search results in the hopes of acquiring a sale. The really ambitious ones hope for more: to acquire a customer. The stellar marketing organizations also actively work to retain that customer. Sadly, it is a rapidly narrowing funnel, working through those 3 levels of goals.

Google (and any other search engine, but let’s be real) has to strike a very tight balance: if it overwhelms the results with paid placement, especially paid placement that is strikingly less relevant than organic results, then consumers will balk. Google can make more money in the short term but may erode their ability to make money in the future as consumers turn to other sources in hopes of finding greater relevance.

In fact, it is likely exactly this conflict of interest and the fact that GenAI chat interfaces don’t (yet) have this conflict, that drove consumers to start using ChatGPT and Perplexity for all kinds of search but especially product search, to the point that Google noticed – thus the “red alert” moment.

So Google has been working on an “AI Mode” tab for their search, and has started rolling out explanations to its search business customers as to what’s new and how it’s going to work. Mike King breaks it all down in the article I link to above and I definitely am not going to rehash it, but there are some key things that you need to know about how it works before we can talk about what it means.

A Brief Explanation of Google’s AI Mode

SEO meant retailers had to invest a lot in understanding the terms that consumers used when looking for the kinds of things a retailer or brand sells. It’s been an arms race between Google and its algorithms, and the SEO consultants and analytics tools that deciphered what powered those algorithms and made recommendations on how to game them.

“Game them” is harsh, but also true. You get the behavior you measure, and a brand’s ability to optimize against that next update to the algorithm could be life or death to customer acquisition, and especially to customer acquisition costs (CAC).

With AI Mode, gaming the system is going to be a lot harder. Basically, a consumer enters a prompt – in this case all I care about are the kinds of searches that lead to product purchases. But the LLM now sits between the prompt and the result, not an algorithm. It’s not about scoring high against the terms the consumer uses in her prompt. It’s about understanding how the LLM interprets that prompt and decides what to use as sources to synthesize a response.

When a consumer enters a prompt, say, “Give me ideas of what to wear on a date night this Saturday”, the LLM is going to dissect that question into dozens of sub-questions and related questions, then conduct its search for sources that help it best answer those questions, then it’s going to look across the answers and decide what is the most accurate interpretation of the consumer’s prompt and then build an answer to the original question out of the most relevant sources that answer what the LLM has determined is the most accurate interpretation of the prompt.

That’s bad enough – right now, brands don’t get to see the prompts, and they don’t get to see the fan out of new questions the LLM generates, let alone how their content ranks as relevant for those fan-out questions, let alone how the LLM decides which fan-out questions to weight more heavily in building a response.

But it gets worse. As consumers stay logged in and build a history of prompts and actions they’ve taken in response to the outputs, that history will increasingly influence the LLM’s interpretation of the current prompt. And brands won’t have access to that at all, and even if they had perfect access, their ability to tailor content to individual prompt preferences is as laughable as the ways they “personalized” their websites in the early days of the discipline.

Side note: will there be a future of AI agents that are designed to be synthetic customers so that brands can test how to dynamically generate content based on what they think LLM’s will be looking for? Agents to answer questions from bots? Oh yes. Definitely. When that overwhelms human traffic on the internet, I guess that’s when Guy Pearce appears and shows us all what books are.

Retailers’ Response to Generative Search

Retailers are scrambling to optimize their data to better show up in answers to lifestyle intent questions. Instead of indexing their products on color, material, origin, etc. they are adding in GenAI-style tags like “great for hiking” or “great for date night”, in the hopes of scoring against prompts from consumers asking about what’s good for each of those activities.

But that’s not enough. I scoffed at the idea of trying to “influence AI” but it’s more true than I initially gave it credit: ultimately, you have to be able to divine a whole bunch of questions that could be derived from a prompt and be an easy, ready friend to the AI that is looking for resources that best answer those questions. You have to guess how the LLM reasons – when the LLM is using that customer’s specific prompt and past prompt history (and reactions to that prompt history) as basically a longer term memory of context.

There is no easy answer here, other than whatever you do for SEO is no longer relevant. You need copy and content that answers questions you have to guess that an LLM may ask. Good luck.

But… SEO and Paid Advertising is Big Business

It sure is. And it’s been a really great business for the search companies company monetizing it. The pandemic forced retailers to dump their advertising budgets into digital, and all those dollars chasing what was still a finite amount of consumer attention (more than before, since everyone was home, but not limitless) did what too much demand and not enough supply does: it drove up prices. Specifically, customer acquisition costs (CAC).

At the same time, privacy laws and features made it harder to track and target desirable customer segments. Not only are retailers paying more, they’re getting less. That’s great for the sellers of consumer attention. Not so great for the buyers.

But if LLM’s make it harder to game the system, they also make it harder for search companies to monetize the system. OpenAI isn’t a search company, so they haven’t cared if they disrupt the search industry. Which is why, at the moment, they don’t really care if consumers who use GenAI search summaries are cutting click throughs – real, tangible traffic to a website – by over one-third. So far.

What happens to all that money being poured into search today, either directly or indirectly? This is where the MIT guys going after ChatGPT on a potential future of selling not just intent, but the ability to influence intent becomes interesting/scary. And I think we have to accept that there is going to have to be a way for these companies to monetize their LLMs in a search-oriented world. There still are not enough people willing to pay $20/month, and definitely not enough people willing to pay $200/month, and even if everyone did, it still doesn’t cover the cost of GenAI’s progress, even as that progress gets cheaper and models themselves get cheaper to run.

A New Level of Trust Balancing Act

There’s one more aspect to all of this that we have to throw in the pot: the trust balancing act. To make it plain, I’m going to start with something that will probably date me: checkout coupons. I know they’re still around here and there, but for a brief period there was a real struggle between the cashier who wanted you to wait and please take half a roll of printer paper’s worth of coupons and you wanting to get out the door without grabbing that wad of uselessness.

Aside from the fact that it was a terrible place to try to influence customer behavior – immediately after a sale! – it also struggled with the trust balancing act. As consumers became more aware of all the data that retailers were collecting about them, they started to get actually angry about those stupid coupons – because they weren’t relevant! My personal experience was the amount of money that Pepsi spent on trying to get me to switch from Coca Cola, when a very brief review of my purchase history would easily affirm that they could give it away for free and I wouldn’t take it.

What hurt the retailer (and the coupon companies) was that they were making money in the short term, selling coupon access to customers to anyone who wanted it, instead of giving customers the coupons they actually wanted. That worked in the short term but long term, it eroded customer trust in the retailer. “You’re not looking out for me, you’re selling my data to the highest bidder.”

That’s true for SEO too. The bargain with SEO was that the brand at least put some thought into what would resonate with me as a consumer, so if you put the effort in to rank high, even paid for it, I’m going to give you the benefit of the doubt. And also, as a human I’m lazy and like pretty much everyone else on the planet I don’t go deep in the search results, willing to trade some gamesmanship and monetization of the search algorithms to save having to wade through 20 websites.

In a generated search results world, all bets are off. It’s opaque to the brands trying to get in front of consumers, and it’s opaque to consumers what brands are doing that get them into the AI Mode summary. That makes the next big question this: how much can GenAI monetize customer intent – and balance giving the consumer what is actually most relevant vs. giving her what someone else is willing to pay to put in front of her.

What If They’re Selling Influence Instead of Access?

But it’s not even “just that simple” (and that’s bad enough). Just looking at how more and more (younger) people are using these tools, I feel like people are particularly vulnerable to being manipulated when it comes to shaping intent and then fulfilling it. As the MIT guys point out, what’s the line between “Do you want to see a movie tonight?” and “You should see a movie tonight”? I feel like this teeters dangerously close to the dilemma posed by checkout coupons. Consumers may not notice at first, but the first time GenAI misses on a recommendation and consumers feel like they’re being manipulated to serve a brand’s objective rather than their own, then trust is broken. And that’s very difficult to rebuild.

Here's a this-week example: Alta Style, brought to my attention by Jason James at work (thanks, JJ!). I’m waiting for a company that is going to offer me an agent for me (versus the retailer’s chatbot). I build context by giving it the things I already own, and then it gives me recommended outfits to wear. Every once in a while I want to ask it to recommend finding a new piece to include.

At first glance, Alta seems to offer this. But Alta starts with a checkbox of brands – most of which I don’t shop (I’m sorry, I missed where Costco was going to fall in between Coach and Michale Kors – both are Aptos clients, but I am not their target customer). I haven’t dropped in images of any clothes that I own yet, and I already feel like this is going to be biased towards these high-end retailers and not to me. But again, how is Alta going to make money? This is not an altruistic effort to make the world more fashionable. My guess: by selling my intent to the highest bidder.

Combine that with long-running issues around CAC and the fact that retailers have historically been terrible at retaining customers – things are about to get way more expensive, while also getting murkier and more disconnected between the actions you take and how those actually end up getting presented to consumers.

What Can Retailers Do?

Ironically, I can see a future where we revert back to the store being the most important discovery channel. I’ve talked about that before in terms of the trust equation: consumers used to go to stores in order to develop trust in a brand such that they would follow you on social channels. But digital disrupted that, so that consumers wanted to get to know your brand in digital channels before they would trust you enough to actually go to a store.

If a consumer really wants to get to know your brand, where are they most likely to get a genuine, unfiltered view? The only place they can get that in a way they can trust that what they see is genuinely your brand is in a physical store. Once they have that more visceral experience, then they can better judge whether a bot or agent or search summary is hitting the mark, and not actually serving someone else’s agenda (or at least does both).

Because, as retailers well know, once trust is broken, it’s very difficult to get back.

What Did We Learn This Week?

This whole article this week is a “what did I learn” so instead I’m going to make one more comment about RetailROI. The charity was started with the intent of having retailers and retail technologists use their connections to provide access to things that small, local charities normally would not be able to get their hands on – think, a commercial kitchen for a school, for example. Or logistics expertise to import something. Aptos once had a warehouse full of obsolete retail-hardened tablets that we stripped down to be able to run the Google Suite, and that became the classroom tablets for a bunch of kids in Haiti.

With the shifts in technology towards more consumer devices, we don’t have that kind of opportunity any longer. I’m not at the same level as a restaurant company exec inspiring a supplier to donate a kitchen, or a grocery store exec inspiring a wholesaler it works with to find and donate seed to some farmers. I sometimes feel a bit inadequate that I can’t make connections like that!

But, modestly, I do have a platform. People ask me all the time why I’m so passionate about retail. My answer is, because if retail didn’t exist, we’d all be at home knitting sweaters or hoeing the fields. I don’t know about you, but I suck at both those things.

People lift themselves out of poverty when they can produce more than they consume, and sell or trade the surplus for things they otherwise wouldn’t be able to get their hands on. Guess what? Retail makes that process more efficient by aggregating access to many things from many different places. And makes it possible for you to be reading this article instead of hooking a horse up to a plow or out hunting deer or something.

Sometimes the things people sell or trade for are depressing and frustrating, like access to health care. That’s what I mean by inspiring and frustrating. The inefficiencies in parts of the world where there are extreme barriers to this very basic exchange – that’s frustrating. The people like the people at Casita Copan and COMFE are inspiring, as are the people who live that frustration every day and don’t just keep going, but try to help others in their community stay afloat too.

So here’s me using my platform: If you can, go on a RetailROI trip. See for yourself parts of the world that you’d never see or understand in this way as a tourist. Find a RetailROI-sponsored charity that can benefit from your connections. Or heck, any charity, for that matter. Support Super Saturday, RetailROI’s most important fundraiser (the Saturday before the NRF Big Show). And also, remember that retail is an essential, critical function of society, and a pathway out of poverty.

Maybe you weren’t expecting that answer. Retail can be shallow and consumerist sometimes. But there is deep meaning to be found here too.

Program note: I’m late getting this one out this week because I just spent 30 hours in airports or on planes getting to Bengaluru to visit the Aptos office there. I should be on time for July 14, but it will be coming to you from India!

Until then!

- Nikki

---

I write this weekly article as much to force myself to make sense of what I read on a daily basis as to say anything about it publicly. I will continue to do so - for free - for as long as I can sustain it.

If you like what you’ve read and you haven’t subscribed, please do! It helps me know I’m saying interesting things.

If you’re already subscribed but you can think of someone who would enjoy reading this, please share the article.

Or better yet, share the whole publication and let them get in on the weekly fun!

Thanks for reading, and thanks for your feedback and support!