11% of Consumers Think They've Already Shopped with AI — Here's Why That Changes Everything - Retail Media Breakfast Club

Loading summary

Transcript14 lines

[00:00]
A
Good morning, happy Thursday, and welcome back to the Retail Media Breakfast Club. I'm Kiri Masters and I'm sharing today a snippet with you from one of my favorite podcasts, the Watson Weekly with Rick Watson. And on this episode, the most recent episode that aired, he is talking with Holden Bale, who is the global Chief Strategy Officer at Merkle. And Holden has some interesting, interesting data from Merkle. Both stated preference research from consumers, but also the receipts that Merkle sees across their network. They've been in the data analytics game for a long, long time and so a very worthy source of this type of data. The thing that really struck me, and I'm going to play a snippet of is, is what consumers think that they've done with AI chat assistance. And it reminds me, my son had this really cute word that he used. He doesn't use it anymore. He's far too grown up now. But he would say, I'm sorry, I misunderheard you, as in I misunderstood and I misheard you. And it was, it's actually a great amalgamation of two terms. I misheard you, I misunderstood you. And that is what consumers are doing with their AI chat assistance. 11% of consumers believe that they have bought something through an AI chat assistant like ChatGPT or Gemini. And as Holden points out in this episode, that cannot possibly be true. They are wrong. They are misunderherding themselves. But Holden makes the case we should still listen to these consumers even though they are wrong and in fact, particularly because they are wrong. Let's listen in.
[02:06]
B
We saw this and I think this is one of your earlier points. We saw this with social commerce for a while where people wouldn't admit or didn't fully understand that halo effect. Do you think something similar is still happening here?
[02:18]
C
Yeah, I think I. So I used to get into arguments with our friends at Forrester and like insider intelligence. And of course I, I drink from the cup of the pundit industrial complex of which you and I are both a part of, right? But they say, well, you know, social commerce doesn't seem to be that big in terms of gmv. And I would say, look, Jay, at Forester, I have data that says that 60% of North American consumers above 40 years old think they've bought something through social commerce in the last four months. Is that true? No, of course it's not true. The gmv, the math doesn't make sense. There's too much commerce on the Internet. But it is their perception that like the equivalent of like the iframe in the year of our Lord 2026, you know, they're clicking through a shoppable tag and they're going to your website. In their mind, they've purchased through social commerce. And it's Conway's law, It's the failures of omnichannel on the Internet. We've maybe succeeded with omnichannel physical versus digital. But ironically now we have a channel management problem in digital where I would have people who are heads of E commerce who would say, I'm not going to spend time assorting a product catalog before all the tools are automated inside of like an Instagram shop because it's going to be a rounding error in my revenue. And the marketer owned the searchal channel and the CMO would say, well, you know, it's really annoying for my like media management agency to do all the shoppable tags, so I'm just not going to do it. And the consumer isn't browsing that way. And I could show you data that shows that if somebody sees a shoppable product inside of something like Pinterest, as opposed to something that doesn't have a shoppable tag, they are more cognitively primed and they're more likely to convert as traffic. But we have very siloed notions on the Internet of Commerce attribution. And I think a similar thing is happening with AI because I could tell you about like the 53% of people who say that they go to a website to purchase after they've done some product discovery and evaluation inside of an AI app like Claude, like, you know, gotten the product carousel, they go to a website. 11% of consumers are telling us they purchase inside of the AI app. No, they don't. No, they don't. They don't. They literally don't. Because the math isn't there and the GMV of these apps aren't high enough. And most of them don't have the connections to do embedded checkout. Like, you know, people talk about OpenAI walking away from commerce, which I think, I think they're walking away from parts of commerce, but the consumer feels that they have. And there's a little bit of like arrogance almost in on the brand side to say, like, well, who are you to say? Because they checked out on your dot com, but their perception is that their source of travel was through an AI app. And I think AI is exposing the exact same thing that social exposed, which is if a consumer thinks that they're shopping through a thing, they're shopping through a thing.
[04:49]
A
Wow. I just love this framing. It is such an interesting way to think about it. So few consumers have actually used AI models to complete a transaction. They barely had a chance even when it was available through instant checkout on ChatGPT. But Holden makes the point that it doesn't matter. The consumers think that they did and they are attributing this purchase journey to AI. And all that means is if and when the checkout capabilities, the transaction capabilities come back again. There is this halo effect that's already in place where consumers think that they have already been shopping through this mechanism and they're more open to it. Poor audience targeting is frustrating, but for retail media teams it can be costly too. With GrowthLoop's composable commerce media solution, you can turn your first party data into hundreds of high value audience segments and launch campaigns faster. After partnering with Growth Loop, Instant Commerce pioneer GoPuff scaled from 100 syndicated audience segments to more than 600 and now it takes less than 48 hours to turn around a custom segment for one of their brand partners. Learn more about how GrowthLoop is powering GoPuff's best in class retail media operations and at go.growthloop.combreakfast that is go growth loop.combreakfast the next thing that I wanted to share from the conversation is how the context that these LLMs have about us as consumers can change over time. And the question of how are these horizontal AI assistants going to build up the requisite amount of shopping context and order history that someone like an Amazon has, for example. That is a big barrier to the use of these models because we want personalization. We want to be remembered by our AI assistants. We want our preferences remembered. We want the ages of our children remembered. But like also not in a creepy way. And this is another really interesting part of the conversation where Rick and Holden talk about what it is going to take for horizontal AI agents to actually collect all of that information about us. Let's listen.
[07:41]
C
The most obvious thing is a replenishment trip mission. I need Dove body wash. I'm out. I need detergent. I always get the same detergent. My socks have holes in them. I buy Hanes or whatever. We estimate if you could, which you can't really, but if you could code the share of wallet of basically the entire Internet into trip missions on the B2C side, at least 25% would be replenishment. And that's the obvious stuff. And that's the stuff where you can imagine over time. I'm training a personal AI assistant with a tremendous amount of context over time about my life.
[08:13]
B
Can I stop you, Holden? Who the hell has the time? Like, who's got that kind of time with, like, working families and all these things? Like, who the, who are these people who have time to train their AI, like in having worked on AI of on my own. It's like the context is the thing.
[08:29]
C
The context is the thing.
[08:30]
B
The context is the thing because the model can do whatever and we already have way more model power than we probably need. It's the context that nobody wants to sit there all day and tell them, like, oh, this person named Joan is my aunt and she met my mom in this time and like 20 years ago. And that's why we buy this soap.
[08:50]
C
Right?
[08:50]
B
You know what I mean? That's the answer. It's not because we like the soap.
[08:53]
C
Well, nobody needs the soap. What you need is one thing the model doesn't have, which is why it's important they come to term with marketplaces, which is the vast majority of the GMV on the Internet flows through is I need your order history. That is the one piece of context data that I have if I'm a marketplace or if I'm a retailer and I don't have if I'm an LLM app. On the flip side, why did they give that up?
[09:15]
B
Like, you're saying, like, I understand the OTAs and travel and like, you know, I'm going to go short Travelocity, if that's still a thing anymore. I haven't been on Travelocity in like, probably 15 years. But Amazon's not going to give it up. And that's. That's 50% of the US right there.
[09:30]
C
Create their own discreet AI app that's embedded into their own retailer experience or that it's a standalone thing, whether it's Rufus or the artist formerly known as Rufus, that is now Alexa. Again, the argument you could make that they might, that they might consider doing this is I and you and the people with working families are just going to spend more and more time talking to our AI app and it's just going to become the. We go to when 87, 76% of the time we went to Google to search, we're going to go to our AI app to search and it's going to maintain stateful information and it's going to learn over time. And even the people who are really busy, I'm relatively busy. I transparently use chat GPT to help plan my mom's elder care, because I wanted to ask good questions to lawyers. You're like, what does this have to do with commerce? I'm like, if I trust, if I trust over time this app to help me with something that serious. And I say like, look, I got three grand. Go find me a tv. You think I'm not going to trust that thing? And if that's the first place for product discovery, the companies who choose not to integrate with it, but own a lot of the GMV on the Internet are seeding a very important strategic moment that a new wave of firms could come into.
[10:46]
A
Now I'm going to preview something that I've been working on that I'm really excited to share next week on the show. I have been experimenting with exporting all of my Amazon purchase history into an LLM and seeing how that changes and seeing what those mechanisms look like. So stay tuned next week. There is more to come here, but I do encourage you to go and listen to the full conversation on the Watson Weekly. It's one of my favorite podcasts and this was an excellent interview. I'm hoping to get a hold of the research that I Holden mentioned and that is it from me today. I hope you have a great Thursday and I'll catch you next week.

Summary5 min read

Retail Media Breakfast Club — Episode Summary
11% of Consumers Think They've Already Shopped with AI — Here's Why That Changes Everything
Date: August 6, 2026
Host: Kiri Masters
Guests: Rick Watson (Watson Weekly), Holden Bale (Global Chief Strategy Officer, Merkle)

Episode Overview

This fast-paced, insight-packed 10-minute episode explores a surprising consumer perception: 11% of consumers believe they have already purchased via an AI chat assistant like ChatGPT or Gemini, even though the current technology doesn’t actually support this for most transactions. Host Kiri Masters curates expert commentary from a Watson Weekly episode featuring Rick Watson and Holden Bale, examining not only the data, but also the deeper implications for retail media, consumer behavior, and the future of AI-powered commerce.

Key Discussion Points & Insights

1. “Misunderherding” & Consumer Perception of AI Shopping

Consumer Confusion: 11% of consumers think they've completed a purchase via an AI assistant, even though this isn’t technically possible at scale yet.
Kiri introduces the concept of “misunderherding”—a playful take on misunderstanding and mishearing, reflecting how consumers misattribute their purchase journeys with AI.

“They are misunderherding themselves. But Holden makes the case we should still listen to these consumers even though they are wrong, and in fact, particularly because they are wrong.” — Kiri Masters [01:14]
The Power of Perception: Holden Bale argues that even if the belief is incorrect, it matters greatly because future behaviors and brand strategies are shaped by perceived experiences, not just objective reality.

“If a consumer thinks that they're shopping through a thing, they're shopping through a thing.” — Holden Bale [04:39]

2. Halo Effects: Lessons from Social Commerce

Parallel with Social Commerce:
Holden recounts that, much like now with AI, earlier research into social commerce (shopping via platforms like Instagram or Pinterest) revealed consumer confusion about when and where a purchase took place.

“I have data that says that 60% of North American consumers above 40 years old think they've bought something through social commerce… Is that true? No, of course it's not… But it is their perception…” — Holden Bale [02:36]
Channel Attribution Challenges:
Brands often have a siloed view of commerce channels, leading to attribution difficulties.

“We have very siloed notions on the Internet of Commerce attribution. And I think a similar thing is happening with AI.” — Holden Bale [03:42]

3. The Implications for Retail Media Strategy

Building Readiness for AI Commerce:
When/if AI chat apps do enable seamless checkout, a “halo effect” means many consumers will feel like they’ve already done it, making them more open and ready to use these tools for shopping.

“All that means is if and when the checkout capabilities, the transaction capabilities come back again. There is this halo effect that's already in place where consumers think that they have already been shopping through this mechanism and they're more open to it.” — Kiri Masters [04:54]

4. The Context Challenge for AI Assistants

Personalization and Order History:
AI assistants don’t possess the detailed purchase history that platforms like Amazon have, making deep personalization a challenge.

“The context is the thing because the model can do whatever and we already have way more model power than we probably need. It's the context that nobody wants to sit there all day and tell them…” — Rick Watson [08:30]
Replenishment Missions:
A significant portion (estimated 25%) of B2C online shopping is replenishment trips (routine buying of the same products), where accumulated context and history are most essential.

“We estimate…at least 25% would be replenishment. And that's the obvious stuff…the stuff where you can imagine over time. I'm training a personal AI assistant with a tremendous amount of context...” — Holden Bale [07:48]
Marketplace Leverage:
Marketplaces are unlikely to share their data with third-party LLMs, making it hard for external AI tools to match the level of contextual personalization consumers expect.

“That is the one piece of context data that I have if I'm a marketplace or if I'm a retailer and I don't have if I'm an LLM app.” — Holden Bale [08:53]

“Amazon's not going to give it up. And that's. That's 50% of the US right there.” — Rick Watson [09:15]

5. Future Directions: AI as the Consumer’s First Stop

Shift in Discovery:
As consumers rely more on AI assistants for initial discovery, retailers who don’t integrate with these platforms risk losing strategic relevance.

“...if I trust over time this app to help me with something that serious. And I say like, look, I got three grand. Go find me a tv. You think I'm not going to trust that thing? And if that's the first place for product discovery, the companies who choose not to integrate with it...are seeding a very important strategic moment…” — Holden Bale [09:37]

Notable Quotes & Memorable Moments (with Timestamps)

On Perception Defining Reality:
“If a consumer thinks that they're shopping through a thing, they're shopping through a thing.”
— Holden Bale [04:39]
On Attribution Confusion:
“There’s too much commerce on the Internet. But it is their perception that...they're going to your website. In their mind, they've purchased through social commerce.”
— Holden Bale [02:40]
On Context as the Missing Ingredient in AI Shopping:
“The context is the thing because the model can do whatever and we already have way more model power than we probably need. It's the context that nobody wants to sit there all day and tell them...”
— Rick Watson [08:30]
On Marketplaces’ Reluctance:
“Amazon's not going to give it up. And that's. That's 50% of the US right there.”
— Rick Watson [09:15]

Key Timestamps

[00:00–02:05] — Show intro, Kiri’s framing of the AI shopping myth, and introduction to the guest snippet
[02:05–04:49] — Social commerce vs. AI commerce: Perception, attribution, and the power of “halo” effects
[04:49–07:40] — Kiri’s analysis: Why misunderstood consumer journeys still matter for marketers
[07:40–10:46] — The challenge of AI assistant context, order history, personalization, and what it takes for LLMs to compete with retailers/marketplaces

Summary Takeaway

This episode highlights a critical insight for retail media professionals: consumer perception has a profound impact—sometimes more so than the actual technology available. As AI assistants become more deeply woven into the shopping journey, marketers must recognize and respond to the “halo effect” of AI-driven commerce, even while technical and data limitations persist. Preparing for a near future where AI is the primary gateway to e-commerce is not just prescient—it’s urgent.

Listen to the full episode for more nuanced insights and stay tuned for next week's hands-on experiments with Amazon purchase history and LLMs, as teased by Kiri.

Loading summary

Transcript14 lines

[00:00]
A
Good morning, happy Thursday, and welcome back to the Retail Media Breakfast Club. I'm Kiri Masters and I'm sharing today a snippet with you from one of my favorite podcasts, the Watson Weekly with Rick Watson. And on this episode, the most recent episode that aired, he is talking with Holden Bale, who is the global Chief Strategy Officer at Merkle. And Holden has some interesting, interesting data from Merkle. Both stated preference research from consumers, but also the receipts that Merkle sees across their network. They've been in the data analytics game for a long, long time and so a very worthy source of this type of data. The thing that really struck me, and I'm going to play a snippet of is, is what consumers think that they've done with AI chat assistance. And it reminds me, my son had this really cute word that he used. He doesn't use it anymore. He's far too grown up now. But he would say, I'm sorry, I misunderheard you, as in I misunderstood and I misheard you. And it was, it's actually a great amalgamation of two terms. I misheard you, I misunderstood you. And that is what consumers are doing with their AI chat assistance. 11% of consumers believe that they have bought something through an AI chat assistant like ChatGPT or Gemini. And as Holden points out in this episode, that cannot possibly be true. They are wrong. They are misunderherding themselves. But Holden makes the case we should still listen to these consumers even though they are wrong and in fact, particularly because they are wrong. Let's listen in.
[02:06]
B
We saw this and I think this is one of your earlier points. We saw this with social commerce for a while where people wouldn't admit or didn't fully understand that halo effect. Do you think something similar is still happening here?
[02:18]
C
Yeah, I think I. So I used to get into arguments with our friends at Forrester and like insider intelligence. And of course I, I drink from the cup of the pundit industrial complex of which you and I are both a part of, right? But they say, well, you know, social commerce doesn't seem to be that big in terms of gmv. And I would say, look, Jay, at Forester, I have data that says that 60% of North American consumers above 40 years old think they've bought something through social commerce in the last four months. Is that true? No, of course it's not true. The gmv, the math doesn't make sense. There's too much commerce on the Internet. But it is their perception that like the equivalent of like the iframe in the year of our Lord 2026, you know, they're clicking through a shoppable tag and they're going to your website. In their mind, they've purchased through social commerce. And it's Conway's law, It's the failures of omnichannel on the Internet. We've maybe succeeded with omnichannel physical versus digital. But ironically now we have a channel management problem in digital where I would have people who are heads of E commerce who would say, I'm not going to spend time assorting a product catalog before all the tools are automated inside of like an Instagram shop because it's going to be a rounding error in my revenue. And the marketer owned the searchal channel and the CMO would say, well, you know, it's really annoying for my like media management agency to do all the shoppable tags, so I'm just not going to do it. And the consumer isn't browsing that way. And I could show you data that shows that if somebody sees a shoppable product inside of something like Pinterest, as opposed to something that doesn't have a shoppable tag, they are more cognitively primed and they're more likely to convert as traffic. But we have very siloed notions on the Internet of Commerce attribution. And I think a similar thing is happening with AI because I could tell you about like the 53% of people who say that they go to a website to purchase after they've done some product discovery and evaluation inside of an AI app like Claude, like, you know, gotten the product carousel, they go to a website. 11% of consumers are telling us they purchase inside of the AI app. No, they don't. No, they don't. They don't. They literally don't. Because the math isn't there and the GMV of these apps aren't high enough. And most of them don't have the connections to do embedded checkout. Like, you know, people talk about OpenAI walking away from commerce, which I think, I think they're walking away from parts of commerce, but the consumer feels that they have. And there's a little bit of like arrogance almost in on the brand side to say, like, well, who are you to say? Because they checked out on your dot com, but their perception is that their source of travel was through an AI app. And I think AI is exposing the exact same thing that social exposed, which is if a consumer thinks that they're shopping through a thing, they're shopping through a thing.
[04:49]
A
Wow. I just love this framing. It is such an interesting way to think about it. So few consumers have actually used AI models to complete a transaction. They barely had a chance even when it was available through instant checkout on ChatGPT. But Holden makes the point that it doesn't matter. The consumers think that they did and they are attributing this purchase journey to AI. And all that means is if and when the checkout capabilities, the transaction capabilities come back again. There is this halo effect that's already in place where consumers think that they have already been shopping through this mechanism and they're more open to it. Poor audience targeting is frustrating, but for retail media teams it can be costly too. With GrowthLoop's composable commerce media solution, you can turn your first party data into hundreds of high value audience segments and launch campaigns faster. After partnering with Growth Loop, Instant Commerce pioneer GoPuff scaled from 100 syndicated audience segments to more than 600 and now it takes less than 48 hours to turn around a custom segment for one of their brand partners. Learn more about how GrowthLoop is powering GoPuff's best in class retail media operations and at go.growthloop.combreakfast that is go growth loop.combreakfast the next thing that I wanted to share from the conversation is how the context that these LLMs have about us as consumers can change over time. And the question of how are these horizontal AI assistants going to build up the requisite amount of shopping context and order history that someone like an Amazon has, for example. That is a big barrier to the use of these models because we want personalization. We want to be remembered by our AI assistants. We want our preferences remembered. We want the ages of our children remembered. But like also not in a creepy way. And this is another really interesting part of the conversation where Rick and Holden talk about what it is going to take for horizontal AI agents to actually collect all of that information about us. Let's listen.
[07:41]
C
The most obvious thing is a replenishment trip mission. I need Dove body wash. I'm out. I need detergent. I always get the same detergent. My socks have holes in them. I buy Hanes or whatever. We estimate if you could, which you can't really, but if you could code the share of wallet of basically the entire Internet into trip missions on the B2C side, at least 25% would be replenishment. And that's the obvious stuff. And that's the stuff where you can imagine over time. I'm training a personal AI assistant with a tremendous amount of context over time about my life.
[08:13]
B
Can I stop you, Holden? Who the hell has the time? Like, who's got that kind of time with, like, working families and all these things? Like, who the, who are these people who have time to train their AI, like in having worked on AI of on my own. It's like the context is the thing.
[08:29]
C
The context is the thing.
[08:30]
B
The context is the thing because the model can do whatever and we already have way more model power than we probably need. It's the context that nobody wants to sit there all day and tell them, like, oh, this person named Joan is my aunt and she met my mom in this time and like 20 years ago. And that's why we buy this soap.
[08:50]
C
Right?
[08:50]
B
You know what I mean? That's the answer. It's not because we like the soap.
[08:53]
C
Well, nobody needs the soap. What you need is one thing the model doesn't have, which is why it's important they come to term with marketplaces, which is the vast majority of the GMV on the Internet flows through is I need your order history. That is the one piece of context data that I have if I'm a marketplace or if I'm a retailer and I don't have if I'm an LLM app. On the flip side, why did they give that up?
[09:15]
B
Like, you're saying, like, I understand the OTAs and travel and like, you know, I'm going to go short Travelocity, if that's still a thing anymore. I haven't been on Travelocity in like, probably 15 years. But Amazon's not going to give it up. And that's. That's 50% of the US right there.
[09:30]
C
Create their own discreet AI app that's embedded into their own retailer experience or that it's a standalone thing, whether it's Rufus or the artist formerly known as Rufus, that is now Alexa. Again, the argument you could make that they might, that they might consider doing this is I and you and the people with working families are just going to spend more and more time talking to our AI app and it's just going to become the. We go to when 87, 76% of the time we went to Google to search, we're going to go to our AI app to search and it's going to maintain stateful information and it's going to learn over time. And even the people who are really busy, I'm relatively busy. I transparently use chat GPT to help plan my mom's elder care, because I wanted to ask good questions to lawyers. You're like, what does this have to do with commerce? I'm like, if I trust, if I trust over time this app to help me with something that serious. And I say like, look, I got three grand. Go find me a tv. You think I'm not going to trust that thing? And if that's the first place for product discovery, the companies who choose not to integrate with it, but own a lot of the GMV on the Internet are seeding a very important strategic moment that a new wave of firms could come into.
[10:46]
A
Now I'm going to preview something that I've been working on that I'm really excited to share next week on the show. I have been experimenting with exporting all of my Amazon purchase history into an LLM and seeing how that changes and seeing what those mechanisms look like. So stay tuned next week. There is more to come here, but I do encourage you to go and listen to the full conversation on the Watson Weekly. It's one of my favorite podcasts and this was an excellent interview. I'm hoping to get a hold of the research that I Holden mentioned and that is it from me today. I hope you have a great Thursday and I'll catch you next week.