Episode 161: The State of AI with eMarketer’s Nate Elliott

Summary

Marketecture Podcast Episode 161: The State of AI with eMarketer’s Nate Elliott

Date: February 20, 2026
Host: Ari Paparo (A), with Eric Franchi (C)
Guest: Nate Elliott (D), Principal Analyst and Head of AI Research at eMarketer

---

Overview

In this episode, Ari Paparo and Eric Franchi dive deep with Nate Elliott, eMarketer’s head of AI research, on the rapidly evolving role of AI in marketing, advertising, and commerce. Drawing from 20+ years of digital marketing experience, including stints at DoubleClick and Forrester, Nate provides nuanced, sometimes skeptical, and always data-conscious perspectives on how AI is shaping both consumer behavior and enterprise marketing strategies. The conversation covers consumer adoption curves, the confusion in AI terminology, how AI is (and isn't) influencing buying decisions, the threats and opportunities for publishers, the state of “AEO” (AI Engine Optimization), and the real risks/realities for martech and ad tech.

---

Key Discussion Points & Insights

[05:17] Getting to Know Nate Elliott and His Perspective

• Nate joined eMarketer six months ago, after a long career in digital research and consulting.
• Leads eMarketer’s coverage on how “AI is changing marketing, advertising, and commerce.”
• Jokingly described his job as “trying to sort through all the confusing data on AI” (07:02).
  • Quote: “There is so much bad data that leads to bad insight and bad recommendations... If you guessed a number between 0 and 100 in terms of how many people use AI, I could find a study to back you up.” – Nate Elliott [08:47]

---

[10:31] AI Taxonomy: Why the Term ‘AI’ Is Increasingly Meaningless

• The hosts and guest agree that “AI” encompasses a vast range of technologies and use cases, leading to constant crossed wires in industry conversations.
• Nate’s Take: It’s his job now to play “taxonomy cop” to help bring seriousness and specificity to discussions.
  • Distinctions include “AI for productivity” (employee tools), enterprise-level AI (efficiency/effectiveness), and AI-powered products.
  • Quote: “We use these two letters [AI] as a monolith... In practice, these are worlds apart.” – Nate Elliott [11:27]
• Joking that “asking the AI to create the taxonomy” might soon be the industry’s solution.

---

[12:22] Consumer Adoption: Are We All AI Users Now?

• Passive vs. Active Use: The distinction between involuntarily using AI (everyone, via Google/Netflix/feeds) versus seeking it out (like ChatGPT).
  • Only a minority are “active” users—roughly 15-20% of US online users engage weekly with tools like ChatGPT (15:03).
  • “100% of people use AI whether they know it or not” (15:44); but fundamental behavior change is more limited.

[14:10] Google’s Messy AI Branding and Measurement Issues

• Google’s “branding mess” with overlapping products and unclear statistics creates further confusion about real adoption and value.
• Microsoft’s “Copilot” brand praised as more coherent.

---

[16:13] Agentic AI and Shopping: Is Anyone Really Buying?

• True agentic transactions (“buying via ChatGPT”) are minuscule—forecasted to be << 1% of e-commerce even by 2026.
• Influence is much broader, but hard to measure: if you count every AI-powered recommendation, you can claim 100% of online shopping is “influenced by AI.”
• eMarketer’s approach: Measures only direct prompt-to-purchase in the same session for clarity.

---

[19:47] AI-Driven Discovery, “AEO,” and the Marketer’s Dilemma

• Marketers are extremely concerned about being included in chatbot/AI answers (“AI Engine Optimization”/AEO).
• The parallel with SEO in the late ‘90s: “It was a disaster, because we didn’t know anything... The reckoning that happened in SEO is going to come to AEO, too.” – Nate Elliott [20:37]
• Wild West tactics abound, reminiscent of early SEO “gaming the engines.”
• Quote: “We still don’t know if we can accurately measure how often these things are mentioning different products.” – Nate Elliott [21:26]

---

[22:01] Paid Media, Measurement, and the Hope for Better Data

• The promise of ad platforms for AI engines: not because of ads per se, but for the keyword tools and campaign planning data that would benefit both paid and organic optimization.
• Quote: “I’m really excited about ads in AI engines...hoping that the AI platforms that sell ads will start to offer the same kind of keyword planning tools that search engines have offered for decades.” – Nate Elliott [22:16]

---

[23:20] What Are CMOs and Agencies Actually Doing with AI?

• CMOs focus on basic AEO and increasing team productivity rather than much else.
• Agencies are fixated on AI as both threat and tool; incentive to automate what they do before clients or competitors do for them.
• Huge tension: Real adoption is slowly incremental, despite bombastic headlines about “the SaaS apocalypse” or “95% of marketing” being automated.
• Machine learning already underpins martech and adtech tools; what’s new is language, not fundamental capability—“more evolution than revolution” (24:48).
• Quote: “Saying that 95% of marketing will be AI is just nonsensical to me.” – Nate Elliott [29:06]

---

[27:26] AI's Impact on Workforce and Tools

• Many managers and CEOs hope for cost savings from AI tools, but will likely face significant “upheaval” before realizing actual business change.
  • “You need people to tell the tools what to do. At least at this point.”

---

[31:07] Labelling and Disclosure in Advertising

• Skepticism about looming regulatory and industry frameworks for labelling AI-generated content.
• Quote: “I don’t know why we need to label AI generated content and advertising, especially given that 100% of what’s coming out of agencies...will have used AI at some point.” – Nate Elliott [31:07]

---

[36:22] The Content Marketplace Debate

• Industry wrangling over whether large AI companies will be “forced” to pay publishers for data and content used in LLM training and answer generation.
• Amazon and Microsoft are creating content marketplaces to ensure publishers get paid (and to assure enterprise cloud clients).
  • Microsoft reportedly has already spent ~$10 million.
• Google’s claim that scraping for AI is equivalent to web search is meeting increasing skepticism.
• Quote: “AI overviews in particular seem less likely than other LLM services to offer people a chance to actually click through and go find the source of the content.” – Nate Elliott [36:49]
• Google and Microsoft have a business incentive to maintain publisher health, unlike fast-moving AI startups.

---

[41:54] Existential Questions for Publishers (and AI Platforms)

• If content quality/value erodes, both search and AI engines lose value.
• “If this content goes away, then both search and AI die. That’s the bottom line for Google.” – Nate Elliott [40:56]
• Google is “somewhat trustable as the adult in the room,” but if OpenAI “wins the race,” industry norms may further erode.

---

[43:37] The Rise of “REO” (Reddit Engine Optimization) and Content Floods

• With Reddit’s newfound status as favored LLM input (and due to paywalls blocking other major sources), brands are flooding Reddit with content.
• Reddit benefits with better ad sales, but struggles to package this for advertisers.
• LLMs are easily “influenced” by a single page or manipulated content (echoes of “Google bombing”).
  • Experiment: Publish a listicle or opinionated article, and—thanks to the lack of established inputs—the LLM might confidently repeat its claims as fact.
• Quote: "The ability to get a really confident sounding answer from an AI based on a single, willfully incorrect webpage seems a little bit startling right now.“ – Nate Elliott [45:38]

---

[47:38] WebMCP and Designing the Web for Bots

• Chrome’s proposed WebMCP standard would let sites embed content “for agents, not humans”—enabling companies to optimize for chatbot and agentic access directly.
• Still, data shows enormous divergence between top SEO results and top chatbot responses (as low as 8% overlap!).
• Quote: “If you’re a top 10 [Google] link...and you only have an 8% chance of showing up in LLM responses...that doesn’t sound like good SEO is good GEO.” – Nate Elliott [48:34]

---

[49:30] Out-of-Home Advertising and Silicon Valley Buzz

• Discussion of Vibe’s “targeted” billboard campaigns in San Francisco and the history/impact of strategic real-world advertising for tech brands.

---

[54:30] Agency Rebates and Opacity in Media Buying

• Indie agency Acadia launches insureRebate.com, calling out opaque rebate practices at holding company agencies.
• Jared Belsky claims rebates and non-transparent incentives represent 3–10% of the $600B market and as much as 30% of agency profits.
• General agreement: It’s a recurring scandal the industry seems unable/unwilling to fix, but transparency/earned media may help.

---

[56:40] Amazon Becomes #1 Fortune 100 Company

• Amazon surpasses Walmart to top the Fortune 100 by revenue ($717B).

---

Notable Quotes & Memorable Moments

• “[With AI], If you guess a number between 0 and 100 on usage, I can find a study to back you up.” – Nate Elliott [08:47]
• “We use these two letters [AI] as a monolith… but in practice, you’ve got… worlds apart.” – Nate Elliott [11:27]
• “It really is the fastest technology adoption curve we’ve ever seen.” – Nate Elliott [15:44]
• “If this is search, we’re in 1998. Think about what SEO looked like in 1998. It was a disaster!” – Nate Elliott [20:06]
• “Saying 95% of marketing will be AI is just nonsensical to me.” – Nate Elliott [29:06]
• “100% of what comes out of the agency at some point will be AI powered in some way.” – Nate Elliott [29:06]
• “The reckoning that happened in SEO in the early 2000s is going to happen in the AI engines in the next couple of years.” – Nate Elliott [20:37]
• “You need people to tell the tools what to do. At least at this point.” – Nate Elliott [27:39]
• “If this content goes away, then both search and AI die. That’s the bottom line for Google.” – Nate Elliott [40:56]

---

Timestamps for Key Segments

• [05:17] – Nate Elliott’s background and role at eMarketer
• [08:47] – Problematic AI data and industry confusion
• [10:31] – Taxonomy confusion around “AI”
• [12:22] – Consumer adoption: passive vs. active usage
• [14:10] – Google’s AI branding and usage stats
• [15:03] – Real numbers on active AI users
• [16:13] – Buying things with AI: reality vs. hype
• [19:47] – AEO (“AI Engine Optimization”): Marketer fears and opportunities
• [22:01] – Paid ads and keyword data for AI answers
• [23:20] – CMO/agency reality on AI adoption
• [31:07] – Disclosure requirements and labeling AI content
• [36:22] – Content marketplaces and publisher compensation
• [41:54] – Existential threats for web publishers
• [43:37] – The “Reddit effect” on LLMs and “REO”
• [47:38] – WebMCP and agent-specific web content
• [49:30] – Out-of-home advertising and Silicon Valley
• [54:30] – Agency rebates and transparency debates
• [56:40] – Amazon becomes #1 Fortune 100 company

---

Conclusion

This episode provides an insightful, data-driven, and occasionally skeptical overview of the state of AI’s impact on marketing, commerce, and media. Nate Elliott reveals the deep confusion and hype in the current landscape (“taxonomy cop required”—[11:27]), guides listeners through the actual numbers on AI adoption and shopping (“15–20% weekly users”) and highlights both the practical and philosophical challenges marketers, publishers, and agencies face as AI increasingly permeates every layer of the digital economy.

For more from Nate Elliott, check out eMarketer’s “Behind the Numbers” podcast and research at eMarketer.com.

Transcript

[A]

This podcast is brought to you by MediaOcean. Everyone knows the lumascape. Now behold the MediaOceanScape. A singular graphic showing all the companies they've come together to form the MeteOcean of today. 28 acquisitions in 59 years. Yes, you heard that right. Founded in 1967, they are the OG of adtech. Next year, MeteOcean will be celebrating its 60th anniversary. And they're showing no signs of slowing down with with the acquisition of Innovid and launch of a new AI orchestrator. To learn more, go to mediaocean.comscape that's mediaocean.comscape.

[B]

This episode is brought to you by indeed. Stop waiting around for the perfect candidate. Instead, use Indeed sponsored Jobs to find the right people with the right skills fast. It's a simple way to make sure your listing is the first candidate. C. According to Indeed data, sponsored jobs have four times more applicants than non sponsored jobs. So go build your dream team today with Indeed. Get a $75 sponsored job credit@ Indeed.com podcast. Terms and conditions apply.

[A]

Welcome to the Market Tech podcast. This is Ari Paparo. I'm here with Eric Franchi and our guest today is Nate Elliott, who has been an analyst around our space for 20 plus years. Really looks into the consumer behavior, the business behavior. He's covering AI for eMarketer, so hopefully he'll have a lot of insights for us about what's going on here. Eric, I assume you probably see emarketer in every deck you look at. There's gotta be an E marketer chart, right?

[C]

That red and black, it always stands out. I've used it myself for 20 odd years. Absolutely.

[A]

Branding. We also had Ross Bennis on a couple months ago. He was also great. These analysts really get into the heart of matters and AI is a big remit. Covering AI is pretty daunting.

[C]

Yeah, it is. Nate's also very interesting because he's been at this maybe longer than anybody. He was super early at doubleclick, so he not only has the analyst mind, but he's got the mind of somebody that has like literally seen all of the eras of digital marketing and I think brings very unique perspective. So I of course am very interested in getting his thoughts on AI. There's a ton of AI news, getting his unique perspective on this. I think he's going to be great.

[A]

Yeah. I'm going to bring up the time I first met him and he grilled me as a young product manager at DoubleClick. It was very unpleasant.

[D]

Oh, Man, I can't wait.

[A]

Okay, so we have three weeks until marketecture Live, our big event. So we have announced the startups in the Startup Showcase. So that's pretty exciting. So we had what, 45 submissions? Something like that?

[C]

More than that. We had more than 50. And then we had some stragglers that we had to turn away virtually because they missed the deadline. So similar to last time, I think we're in that, you know, almost 60 range. So we picked our top five.

[A]

Okay, run through it. Tell us what are the five.

[C]

Absolutely. So again, the theme of the conference is consumer and control. So we really wanted to bring out companies that are kind of pushing the envelope on, like, new formats, new spaces, using data in unique ways. So the finalists, in alphabetical order are Crowd Access, which is measurement for experiential marketing.

[A]

Interesting.

[C]

So a completely kind of new and adjacent space. Data Links, which is data readiness for enterprise. AI Gamera, AI Powered inventory quality platform.

[A]

With our competitive podcast host, we gotta give Gareth a little bit of a hard time, you know, going against the king here. Architecture podcast.

[C]

Absolutely. Merchkit, which is interesting. Kind of like AI powered product catalog tools, which is like applications to marketing with commerce being such an important space. And then Open Glass, which is kind of an SSP for unique CTV formats. Trying to push the inventory forward on things like pause ads and all these other unique formats that I think a lot of us are interested in. And right now is a challenge of scale. So five companies doing completely different things. The submissions were all great. It was super tough to pick these five, and I think it's going to be another barn burner.

[A]

That's awesome. And if you weren't at the last one, we give two awards. First of all, no money, no monetary awards. Sorry, guys. We get one award from the experts. So Eric and Sonia will vote on their favorite and then the attendees get to vote. And maybe there'll be the same award, maybe they won't, but you get a nice little piece of loose sight either way. So really looking forward to it. It's my favorite part of the conference. So there's three weeks left. You should buy a ticket market and there are discounts that we've announced here before. I think it's aripod 30 is 30% if you listen to this podcast. All right, let's dive in with Nate Elliott of eMarketer to find out what's going on in the world of AI. All right, we're here with Nate Elliott from eMarketer. Nate, thanks for being here.

[D]

Thanks for having me.

[A]

So you and I have known each other for a really long time. Like, you know, I remember the first meeting we had. I had taken over Dart Motif, which was this piece of shit product that DoubleClick had for rich media. And you were, didn't you build that product? No. Didn't you build. So you come in as the analyst for I guess for Forester Jupiter at the time and I guess I didn't really know the history that you had been at DoubleClick and had built a lot of that product. And I was very green, no idea what to expect. And you just grilled the shit out of me.

[D]

It's an analyst's job.

[A]

It's like, what is the actual revenue from this product? Knowing full well, it was like near rounded to zero.

[D]

Yeah, I mean it was also what, 2003, the revenue for most digital products was rounded to zero.

[A]

But Kevin Ryan had said to Wall street he thought it would be a $20 million a year product and it was zero.

[D]

I mean, it might have become that at some point. I don't know. You tell me.

[C]

It had to be.

[A]

Oh yeah, it eventually was. We hit 100 a couple years later.

[D]

So KR was right. Just on the wrong timeline.

[A]

Wrong timeline. So enough of the memory lane. So tell me what you're up to. So you're the principal analyst at E Marketer. Are you like the AI guy? Like, is that fair to say?

[D]

I, yeah, I think I'm our head of AI research. That's, that's what someone told me. So I'm, I'm going to believe them on that. But I joined eMarketer about six months ago. Now since our first meeting when I was at Jupiter, I stayed at Jupyter. We got acquired by Forrester. All told, I was with that company for about 12 and a half years. I ran my own one man band doing research and consulting called 19 Insights for about a decade. And now I've been an E marketer for about six months, leading our coverage of how AI is changing marketing, advertising and commerce.

[A]

So that's a big topic.

[D]

That's definitely doable for one person, right?

[A]

Yeah, it's a question about whether everything in E Marketer becomes AI or if AI replaces E Marketer. Like that's kind of the race condition.

[D]

I mean, yeah, we're hoping that it doesn't, you know, as, as we tell our salespeople to tell our clients, if you wanted a research firm that just gave you the average of everyone's opinion, then by all means use ChatGPT. But if you're hoping that you're even a little bit smarter or that your analysts are even a little bit smarter than the aver, then maybe stick with a real analyst for now.

[A]

Claude, create a graph that's entirely in red with really small footnotes.

[D]

With very long, very small footnotes.

[A]

I love the footnotes. The footnotes are the best part.

[C]

Yeah, don't ever change that. But also Apple Pro, I think of a topic later on in this age of the LLMs needing data, needing so much unique content actually puts companies I think like yours, in a good position.

[D]

Yeah, I mean it's an interesting position for sure. Right. And it's not dissimilar to the conversation that publishers on the open web are having about LLMs and AI chatbots. On the one hand, it could serve as a way to get our ideas or information or content out to a broader audience. On the other hand, if the chatbots are not great about citations and linking and credit, and if the stuff is too freely available through the chatbots, then it creates some problems potentially as well. So we're actively working, as you can imagine, on navigating that situation.

[A]

Yeah, it's really interesting problem for a data rich publisher. So, six months in, what's the vibe? What's the vibe on AI and marketing and sales and advertising?

[D]

Oh man, it's confusing, which is part of why I have a job. So that's good news. I've been in Digital since 97. I started at DoubleClick and I've been an analyst since 2003. So that's almost 23 years now. I've been studying and writing reports and talking about digital advertising and marketing and media and commerce. And I've never seen an industry or a topic as full of data as AI and as full of bad data as people talking about AI. There is so much bad data that leads to bad insight and bad recommendations. And a lot of it is because more people than ever before have the ability to create and publish data. Not everyone knows how to do that very well. And so there's just a lot of general confusion about how many people have actually, actually adopted AI and for what purposes. And you know, as I say to people, if you guess, if you just guessed a number between 0 and 100 in terms of how many people, what percentage of the population uses AI. No matter what number you guess, I can find a study to back you up. And so, you know, trying to sort through that and get to some central truths about how widely this is adopted on the consumer side and how widely it's becoming adopted on the enterprise and agency and marketer side is I think the biggest, biggest challenge for now.

[A]

Yeah, you have all these people talking cross purpose or not communicating where you, you'll hear on a panel that, oh, we would never create a TV commercial using AI. And, and then you hear from an agency head like, we use AI in our creative process extensively and they mean different things. I think.

[D]

Yeah, the taxonomy is, I mean, I was having this conversation at a, at an industry breakfast a couple hours ago, but the taxonomy is frustratingly important in this category right now. When people say AI, it means so many different things. And, and we use these two letters as a monolith to describe us an idea of how technology works. But in practice, you know, you've got AI on the consumer side. How, how are people using and engaging with AI? And even within that there's both, you know, active and passive. There's this idea that every single person who logs online at this point is using AI. They're exposed to AI content or AI powered experiences, whether they choose to engage with those or not. And then if they choose to engage, there are not just hundreds of tools, but all these different ways they can use AI in their personal life and in their work life. And then on the company side, AI means so many different things to so many different companies. And I've broadly started thinking about it as AI for productivity. Right. So AI becomes another tool in our employees toolkit along with Excel and G Suite and Photoshop and everything else. There's AI at the enterprise level informing or becoming the products that we use to improve the efficiency and the effectiveness of the things we do as companies. And then there's AI as products that we're building to sell to people, either services or products. And so when you look across all the things I just described, they're worlds apart. And yet we use these same two letters to describe all of them. And I don't want to be a taxonomy cop, but I think taxonomy cop might very quickly become a pretty big part of my job description.

[A]

Ask the AI to do it. Create the taxonomy.

[C]

Yes.

[A]

Okay, let's, let's try to, let's try to take bite sized chunks. You want to do consumer first?

[D]

Sure.

[A]

All right, so what's going on with consumer? There's a lot of discussion and there have been essays about like this, this huge gaping chasm between early adopters and normies. We're seeing reaction to the AI Bowl, super bowl as being negative. People don't want to hear from these AI things. Meanwhile, people are running their lives. A cloud bot. What's going on with consumers?

[D]

Well, for a small number of people, AI has changed a lot. And for a lot of people, AI has changed a little. And again, we get to this idea that there's no one who uses digital technology in any form who isn't engaging with AI. That's been true for at least a year. Gallup did an interesting study, I think about 14 months ago, where only 36% of people they serve surveyed said that they had engaged with AI powered tools in the previous week. And then it gave people a list of tools and said, have you used these? And 99% had used one of them in the previous week. So this is part of everyone's life right now. But the question of, are people choosing to use these technologies? Is a very different question. And one of the frustrations I have is we're seeing adoption measured on all sorts of different scales. OpenAI reports adoption on a weekly basis. Google reports it on a monthly basis. KPM reports it on a basis of at least once every several months. And I'd argue if you're getting your hair cut or paying your mortgage more often than you're using ChatGPT, you're not really a ChatGPT user. So we're building.

[A]

If I can interrupt for one second, Google's at fault here because Google just shoved AI into the search results and then gives weird statistics about the frequency of usage. And then they'll say publicly, oh, Gemini will never have ads, but their AI results in search have ads.

[D]

So, yeah, I mean, Google's AI products are a branding mess and as far as I can tell, is born of the structure with which they have deployed the AI products. Right. So the AI division of the organization is in charge of Gemini and Nano Banana, and then the search part of the company is in charge of AI overviews and AI mode. And even though they, as far as I can tell, are based on the same large language model, by behind them, they are different services that are run by different parts of the company, have different policies on advertising, have different brands. Somehow Microsoft has done a better job of branding their AI products than Google has.

[A]

Copilot's a good brand. Could you characterize what portion of consumers are, I would call them active AI users, people who have a ChatGPT account and regularly ask it questions.

[D]

Yeah. So depending on the source you look at, if you're defining this as weekly usage, I think we're looking at 15 to 20, 20% of the US online population. Right now. Our survey is going to come back from field in the next few weeks so that we can have our own proprietary number on it. But obviously there are lots of numbers out there. And again, they're all based on different samples and different timescales and all of that. I think 15 to 20% of people are using this stuff at least weekly, and maybe half or two thirds of them are using it daily, which I think is real adaptation, as my colleague Jacob Bourne would call it.

[A]

We're only three years in. Right. So that's kind of a pretty high number.

[D]

It's an amazing number. I mean, it really is the fastest technology adoption curve we've ever seen. And yet you have this very strange situation where 100% of people use AI whether they know it or not. Whether they know or not 50, 60% of people know that they use AI, and yet the percentage of the population for whom this has fundamentally become part of their day to day life is actually relatively small compared to those big numbers and compared to what I think the industry thinks is happening.

[A]

Okay, now let's slice the pie. Really small. Is there any use. I was about to ask what percentage, but like, let me ask, is there any use of agentic technology currently?

[D]

Sure. I mean, honestly, I find the distinction between agentic and generative a little bit boring and reductive at this point. I mean, as far as I can tell, every agentic application involves generative processes and vice versa, and trying to differentiate between them. I mean, I've seen people try to survey consumers on whether they use generative AI versus agentic AI. That doesn't mean anything.

[A]

Has anyone ever bought anything using an AI? Any normal person?

[D]

I don't think any normal person has for sure. I'm certain that transactions have occurred. I think it's exceptionally small. And if what you're referring to is people completing purchases within ChatGPT for instance, or within other agents, then we have forecast that number to be very, very small. We think that maybe 0.1% of all US online retail commerce in 2026 will happen on general purpose chatbots like ChatGPT and Gemini. And even when we go forward, as far as we're willing to forecast that number is below 1%.

[A]

Do you have any data on influence, like how often the AIs are influencing purposes?

[D]

Yeah, so the. We're forecasting that. It's, I mean, it's a. We had some really interesting conversations about this that went on for probably far too long. But we're analysts and we work at research companies. That's what we get to do with our time. And what we realized was if we, if we tried to talk about AI's influence on purchases and didn't limit what that influence, how we, how we define that influence, the number would be 100%.

[C]

Right?

[D]

I mean, Amazon says that Rufus's downstream impact, whatever that means, is about $10 billion per year. But the reality, every time you go to the home screen or homepage of Amazon, you're seeing AI powered product recommendations. So depending on how broadly you define AI's influence, 100% of US online commerce is influenced by AI. The way that we're defining it in our forecasts is either people completing purchases using AI tools like ChatGPT or buy for Me or completing a purchase on a retailer site or app in the same session. So it has to be, I went to ChatGPT, I asked it for the best roll on suitcase that works for European airline sizing restrictions, and I clicked on a link and I bought it in the same session.

[A]

Well, the, clicked on the link is the big gap here because often those links aren't present or not the best links. So people are asking that question and they're going to Amazon and buying something.

[D]

Right. And so I think that we would define that as the same session. I would have to check with our forecasters on that. But it's, I mean, I mean, the idea that you're getting the answer from a chatbot or an LLM and then acting immediately on it is the question that we're trying to ask and the thing that we're trying to forecast. And we actually think that that act of using these tools for product discovery and research and decision making and then purchasing through an established channel like a retailer site or app is the vast majority of how these AI tools are going to influence purchases for the next few years.

[A]

That kind of tees up our transition to talk about marketing, technology and marketing. So aiogeo, the desire for brands to be in the chatbots answers. What's your take on that space and what are you hearing in terms of adoption? Skepticism, value?

[D]

Yeah, I've yet to meet a marketing executive who isn't actively thinking about it and working on it. I've barely met one who isn't terrified of what's happening. And yet, you know, as you pointed out, we're three years in. If this is search, we're in 1998. Think about what SEO looked like in 1998. It was a disaster because we didn't know anything and we were trying to figure out how to, how to track how to measure, nevermind how to influence what showed up and what maybe didn't show up in search results. And a lot of the things that we, we knew could work were not things that ultimately we would consider white hat SEO. And so I think that description of SEO in 1998 is a perfect description of AEO and GEO in 2026. I mean, I was listening into a product demo webcast from one of these AEO geo tracking and optimization companies recently, and they actively and openly used words about, about dominating results and gaming the engines. And I mean, they're basically saying the quiet part out loud in a way that we haven't seen in SEO for decades now. I've no doubt that the kind of reckoning that happened in SEO in the early 2000s is going to happen in the AI engines in the next couple of years. In the meantime, a lot of these companies are trying to make hay while the sun shines. And I'm sure they'll get some, some clients and I'm sure they'll get a lot of VC out of the deal. But it's, I mean, it's very wild west. And the reality is we still don't know if we can accurately measure how often these things are mentioning different products.

[A]

Yeah, because they're also inconsistent. You could ask the exact same question multiple times. Which, which became the case with search. You know, search over time became more and more customized and personalized and very personalized.

[D]

Right.

[A]

But in the early days it was not, I mean, deterministic answer.

[D]

The thing with search that really, I think supercharged SEO was actually the keyword tools that the search engines introduced to help ad buyers. So I'm really excited about ads in AI engines. Not because I necessarily want to see the ads, although they're inevitable, but because I'm hoping that the AI platforms that sell ads will start to offer the same kind of keyword planning tools that the search engines have offered for a couple of decades. For anyone who's not aware, while these tools are designed to help ad buyers figure out how much inventory might be available on different keywords for them to buy, the reality is they're much more valuable to the organic optimizers. And I can see that happening in GEO as well.

[A]

Yeah, that's interesting. I had some people in Twitter kind of mention who's going to be the SEM rush of, of at of AI and things like that. Yeah, ads would bring the money in that would produce more valuable data in a sense. Metadata.

[D]

Absolutely.

[A]

So what else is the CMO or the marketing Leadership freaked out about discovery is obviously a big problem, but what else is on their agenda when they think about their weekly AI status? Meaning?

[D]

Well, it's interesting to me that the CMOs, the marketing executives I've talked to, are not as interested in anything aside from geo.

[A]

They're not interested in other things.

[D]

Well, they're interested in other things, but if you look at how people are using AI and how they might reach people using AI, Geo seems to be the entire conversation. And beyond that they're not thinking about, I mean, they're starting to think about paid ads now because of the ChatGPT experiment, but they're not thinking about the other ways they might leverage AI tools for direct interactions with their audience so much as they're thinking about how they can leverage AI tools internally. And for a CMO that mostly looks like number one, getting their team members to adopt these technologies in their own workflow so that they can improve their efficiency and their effectiveness. And, and then secondarily, I'd say they're looking at sort of enterprise level ways they can use AI powered technologies to improve the things they do at the enterprise level, like ad buying and optimization and measurement.

[A]

Is the SaaS apocalypse going to come to Martech tools?

[D]

Sure. The really interesting thing is machine learning has been a part of all these Martech tools for longer than anyone seems to realize things. We're using different language to describe what's happening and yet what's happening, Is it more of an evolution than a revolution? As far as I can tell? I mean, find me any kind of buy side or sell side exchange or platform that hasn't been using machine learning to inform what it's doing for the past five years or probably longer. And I'll show you a company that probably isn't in business anymore. And so I don't know that we're going to face an apocalypse based on this stuff, but it certainly, I mean, the adoption of these technologies into the platforms companies already use certainly is going to help the marketers, the ad sellers, take another step forward in terms of their efficiency, or at least it should. Good, right? I mean, I still hear a lot of concern about hallucination and incorrect analysis. I've heard people throw around numbers. Over 50% of outputs in an ad buying environment might be hallucinated or inaccurate. That seems wildly high to me, but I think if we can get the inaccuracies, the hallucinations, the problematic outputs down to like, like 10%, 15%, I think the efficiencies that companies will find by Adding more AI and machine learning to the way they buy and sell ads will more than make up for whatever we lose in terms of inaccurate outputs.

[A]

A lot of these martech and ad tech tools have difficult problems to solve that AI would have a hard time with because they collect large amounts of data and process it. And AI doesn't necessarily make any of that easy, easier. Whereas something like content management systems or SEO tools might be quite easy to reproduce.

[D]

I mean, it doesn't make it easier, but it could make it faster. Right. And so if the challenge is I need to find patterns, well, that's exactly what the AI tools are going to be good at helping us with. And if they can help us identify patterns and discard the ones that are not as useful more quickly, then then that is a great potential application. And that's the kind of thing that I'm seeing the companies in the space working on.

[A]

If I can interrupt, if the problem is I need to create a lot of landing pages that collect information from consumers so they can download PDFs. That's a one prompt problem. That's a difficult thing to sustain over time.

[D]

Yeah, it is. And the thing that I think CEOs and some CMOs are hoping for is, you know, is AI application that will allow them to just replace a large part of whatever they're paying for right now, whether it's vendors, agencies, employees, whatever that is. I don't know that that's going to happen. I think a lot of, a lot of CEOs and CMOs are going to end up firing people trying to make it happen. But you know, at the end of the day, these are tools and you need people to tell the tools what you to do. At least at this point, you need people to tell the tools what to do. And so in the next few years, you're going to have a lot of companies going through a lot of upheaval trying to figure out how to use these technologies to change their business, when they haven't even figured out how to use these technologies to make incremental improvements to what they do.

[A]

Right. Which sort of brings up kind of the last topic I wanted to cover, which is the role of agencies. So I think unlike CMOs, agencies are kind of obsessed with AI. It's like the most driving factor in their business right now is how to utilize, how to save money, how to etc. And we've quoted on the show a number of times, as Sam Altman said that he thinks 95% of marketing will be done by AIs so how does that comport with what you're saying the CMOs are thinking?

[D]

Well, so I mean, if you told an agency head that AI could replace 95% of what they do, they would be very excited to try to make that happen on their own so that someone else doesn't make it happen happened to them. And you know, we see Holt Co is talking about the percentage of what they do that they're using AI for. And, and again, I, I go back to the taxonomy question, which is what on earth are you talking about? So I, I don't know a really good way to measure these things in a way that people would want to measure them to have an intelligent conversation. Because if the question is what percentage of your team is using AI? Yeah, we'll get to the point where it's 95%. If we get to what percentage of each person on your team's work are they using AI for? I think depending on the department they're in and the tasks they're given, it could be 100% and it could be 10%. If we talk about what percentage of the ultimate output that the agency is creating, whether that's a media buy or a piece of advertising creative. Is AI influenced or inflected? Well, 100% of what comes out of the agency at some point will, will be AI powered in some way. In the same way we talked about. Basically 100% of people use AI when they go online and 100% of shopping is AI influenced already in some way. But saying that 95% of marketing will be AI is just nonsensical to me.

[A]

Yeah, I have to agree with that. It's sort of a made up number. But one of the big barriers to adoption at the agency level is that they really need to get customer permission for each aspect of AI they want to use on each account. That's at least my understanding. I mean, obviously if you're talking about a TV commercial there, there's gonna be a huge amount of approvals. But the, my understanding is that account teams need to ask and get written permission to do, to use AI for things that are very AI ready, like media planning.

[D]

I mean, I would think that that will work its way into the contracts over time so that that at least removes that friction from the process.

[A]

Process.

[D]

But this is a place where the focus on AI frankly confuses me because the iab and I love what Caroline Kegeritz doing over at the iab, but she just introduced a framework for disclosure. When digital ads were AI generated based on Certain standards. And whether or not the AI adoption or AI usage fundamentally changed the output of the advertising. I don't know why we need to do that. We've used Photoshop to touch up images for 30 years. We've used VFX to fundamentally alter what's happening in video for decades. We're not labeling that stuff. I don't know why we need to label AI generated content and advertising, especially given again, that 100% of what's coming out of agencies in the next few years will have used AI at some point in the process. Maybe it'll be in brainstorming and conceptualization, maybe it'll be in actually making the creatives that people see. But at some point in that process, AI will have touched everything that consumers see. And, and labeling it just seems like an unnecessary nightmare to me.

[A]

I think, correct me if I'm wrong, but the major platforms currently have requirements to label political ads if they use AI, but that's about it. Is that your understanding?

[D]

I don't know the answer to that, actually. So it's one of those things that changes too often.

[A]

Political ads are pretty important, right?

[C]

Sure.

[A]

Because you can make someone say something they wouldn't normally say or do all kinds of bad things with AI.

[D]

Yeah. I mean, one of my challenges with any of these restrictions is that they all seem to rely on the bad actors to do the right thing, which seems like a losing strategy.

[A]

Yeah, it does. All right, that was a really interesting conversation. I appreciate the consumer point of view, which we probably don't talk about enough on this pod. We're going to take a break and come back with. As always, we have a lot of news, a lot of AI news and we're going to talk about A and Geo. So we'll be right back.

[D]

Well, the holidays have come and gone once again, but if you've forgotten to get that special someone in your life a gift. Well, Mint Mobile is extending their holiday offer of half off unlimited wireless. So here's the idea. You get it now, you call it an early present for next year. What do you have to lose? Give it a [e-mail] Switch limited time,

[B]

50% off regular price for new customers. Upfront payment required $45 for 3 months, $90 for 6 month or $180 for 12 month plan taxes and fees. Extra speeds may slow after 50 gigabytes per month when network is busy. See terms.

[C]

All right, we're back everybody with the refresh and guess what? We got a couple of AI stories to talk about. And it's awesome to have you, Nate, because these are things we've talked about on a. On a fairly regular basis. But I think you bring the interesting perspective of talking to CMOs and instead of having so much market data. So first thing is, we talked about this last week. Amazon is supposedly working on its own content marketplace, defined as some sort of marketplace that will pay publishers for either their data that's being sourced in Answers or something like that. So the Journal had, I think, a really good piece that covers this space from a mainstream perspective. A couple things that came out, besides the fact that it seems to be that Amazon is working on something, is how deep Microsoft has been in their efforts for Copilot. So the article quoted Microsoft saying they've invested $10 million to date in their content marketplace, which I think is probably months old, including payments to publishers. So there's something here and they're making a bet here. Ari, what did you put in the notes here from Alan Chappelle?

[A]

Yeah. On the Monopoly Report newsletter, which came out a couple of days ago, Alan makes a negative argument on these content marketplaces, effectively saying there's no forcing function Google, what's going to bring Google to the table, you know, you can't block these scrapers from everywhere, etc. Etc. Which I think is a healthy dose of skepticism. I do think bringing Google to the table is very difficult given that they're continuing to claim that, that the same scraping necessary for search is informing their AI. And you can't separate the two. Meanwhile, folks like Amazon and Microsoft have a positive approach to this because they want to sell the liquidity of this marketplace to their cloud customers. So they're very different perspectives and it is a real question mark about how you get Google to the table.

[C]

Yeah, Nate, I would love for you to talk about this category generally. And so I guess number one, your POV on content marketplaces, especially around how publishers seem to be making this a big bet for getting their business over the line with all the traffic dips. And then number two, have you done any work on this, has Emarketer done any work on this that you might be able to point to as part of this conversation?

[D]

Yeah, we haven't published on this yet, but we're keenly aware of what's happening in this space. And I mean, it's a vitally important question because, you know, the argument that, that Google might make about how the same kind of scraping that they need for search is what they need for AI likely is true, but there's a Big difference in terms of whether people leave AI to go to the sources of that scraped content versus whether they leave search to do that. And it's, you know, I think people underestimate how few people were clicking on the 10 blue links even before AI came along. Even before AI overviews were introduced, it was probably about half, according to the data I've seen of Google. Search results led to a click outside of Google, so that number was already fairly low. AI overviews don't just give people the answers they need without requiring so many clicks, though. They also don't tend to include clickable options. And we're trying to find good data on this right now. But certainly anecdotally, and I've seen people talking about this around the industry, AI overviews in particular seem less likely than other LLM services to offer people a chance to actually click through and go find the source of the content. So that leads to questions around consumer trust in what they're seeing if they can't see the sources. But much more importantly, it leads to real business questions for the publishers. And, you know, right now, AI overviews are by far the number one way that people see AI generated search results results. ChatGPT talks about 800 million monthly user or weekly users rather, but AI overviews were seen by more than 2 billion people last July. That number's absolutely going to be higher right now. And so if those aios are not even offering people the opportunity to click through, then this becomes so much more important than it would be if, if ChatGPT and Claude and Perplexity were winning.

[C]

Do you think this becomes a category content marketplaces like? Do you think in a year or two's time this is a thriving segment of publishing?

[D]

I don't know that it will. I think the lack of a forcing mechanism that you mentioned is a key problem. I mean, I understand the hesitation about whether Google will decide to play absent a forcing mechanism. I will say in their defense, defense when nudged enough, not even forced, but when nudged enough, they have done something approaching the right thing in the past. They have paid a lot of money to publishers for content that appears in News Search. Not nearly enough, I think the publishers would tell you, but certainly much more than zero.

[A]

More than meta, we should also point out.

[D]

Right. So I think there's an inclination for them to do the right thing when pushed in that direction. And I think there's probably a greater understanding from Google and Microsoft certainly than there is from the AI startups that the health of the open web is a necessary precondition for their existence and financial success.

[A]

Yeah, I think there's the cloud angle is important here, which I keep hearing from from folks, which is that enterprises want AI AI capabilities in their cloud providers and enterprises want to know that that capability doesn't have licensing problems and has the most recent data from the web and they're willing to pay for that, or presumably they will be. And so Google might find itself in a corner where Gemini for Google cloud services may not not be as viable a solution as Amazon and Azure solution which have built in licensing.

[C]

Makes sense. The data to fuel the answers. It's just, it's like, it's existential. So I would be surprised if Google and others didn't make some kind of move here because they're dependent on this data for all of these answers to be good such that their services went.

[A]

Yeah, but can you imagine someone in Google saying, you know, the meeting where they're like, well, what stops them from requiring us to pay them on every search result too? Right. You're going from the greatest business model of all time with effectively no cost of sales to opening the door to a massive cost of sale.

[D]

Yeah. I mean, it's not a conversation they want to have. But if this content goes away, then both search and AI die. Right. And that's the bottom line for Google. And I. I'm glad that AI overviews are the primary way people are seeing search results because I think we can somewhat trust Google to be the adult in the room. I mean, if OpenAI actually wins the race for consumer AI adoption, I don't think Sam Altman could care less about this. And I think we'll be looking at a downward spiral. I think again, at least Google and Microsoft and some of the existing platforms and players have been around long enough that they might take the hard decision to start paying for this content, even if it means paying for the content to some extent in search in order to keep their business open and their model viable.

[C]

Well said. Fascinating tension. Let's move on. We'll talk about AEO and geo, which is again another recurring topic. And again, happy to have you on for this, Nate. So New York Times did a fun article, I think, you know, educating the masses on what's going on here. Kitschy title, I don't have it in front of me. But effectively chatbots are the new influencers, meaning you need to influence the chatbots. Two interesting points in this piece. And again, we'll kind of come back to a conversation around state of state. So the state, the obvious which is chatbots are insatiable for detailed information, particularly in areas where there's a void. So brands are publishing a ton more content. They had, I think, the CMO or senior marketer from Athena Health, and she was talking about how they're just publishing like thousands and then tens of thousands, then hundreds of thousands of pieces of content. Because some of these areas, like health and Pharma, I think, are particularly kind of ripe to be the source of record for answers. And then interestingly, they note that paywall content is blocked, but Reddit isn't. And Reddit seems to be a top source for most of the LLMs. So optimizing Reddit, where it's very hard to optimize because you can't control anything as a brand, is becoming super urgent and it's giving rise to a new form of optimization, maybe reo. Right. Reddit engine optimization. So anyway, Nate, I'm fascinated with this category. There's probably a dozen companies we're tracking. They've all raised 20, 30 million bucks. Where does this go?

[D]

Well, the Reddit question's a really interesting one because they are very aware of their status as probably the most important source of information for LLMs, and they seem to be, at least in part, because of that, benefiting financially. Right. We've seen some of the best, well, and also some of the best advertising sales that they've had in their entire existence. And yet I've talked to them and I'm not seeing ad sales packages or AMA packages that include big ad buys or things like that that they can sell to the advertisers who want to try to show up more on Reddit. They talk about strategies, they offer some case studies, a UK car case study where people already loved a certain brand of car and the brand was able to further activate those fans on Reddit. But I mean, if I'm them, I'm absolutely building AMA packages with huge ad buys attached and trying to further benefit from that financially. In terms of the category overall, again, we still don't know how to even measure or track what brands appear in responses, and we still don't know how many people are asking what kinds of questions and what, what categories. And as you said, there are a lot of companies trying to solve this. But one of the most fascinating things I've seen, if you follow Lily Ray, who's a longtime SEO brain who's thinking a lot about how this changes with the advent of AI and AEO and geo, she's done a lot of fun experiments. She's been talking about on LinkedIn where she's publishing intentionally encrypted information in formats that AI seems to like, opinionated articles, listicles, things like that, and able to get the AIs to spit this stuff back out. And because in AI, just as in search, there seems to be a huge percentage of prompts or queries that have never been asked before, it's using Fan out to try to find whatever information it can. And the ability to get a really confident sounding answer from an AI engineering based on a single willfully incorrect webpage seems a little bit startling right now.

[A]

It used back to the days of Google bombing where you would try to find a phrase that had never been used before and you'd put up a blog post and suddenly it would show up. So I have two points on this. The first is anecdote that I had a meeting at Reddit a couple months ago just to say hi and stuff down in the World Trade center office. And I think it was the first time I ever impressed my children with my career. They were like, you went to Reddit? Like, they have a place, there's a place called Reddit. I was like, yes, they have a headquarters. It's not just this, like, it's like, yes, it's not just a website filled with crazy people. They have a location. The second thing I'll say more seriously is we didn't, I don't think we covered it. But this new proposal that's out from, From Chrome called WebMCP, that allows websites to embed functions and data for agents, not for humans, is really important. I think I'm gonna probably write about in my newsletter and it's kind of relevant to this conversation because for the first time, a website could be designed with information specifically for the agents and, and it could be very long, voluminous content that humans wouldn't normally want to absorb. You could of course do this with hidden links and stuff like that, SEO type techniques, but why not just give the agents a full catalog of thousands of megabytes of data about your products if the agent wants it. And I'm trying to connect the dots here between special formats for the web versus agent reading versus, versus what we're talking about. Aeo.

[D]

Yeah. Google themselves will tell you up and down that good SEO is good geo and you should just do the same stuff you've been doing for good ethical SEO for two decades is what you should be doing to appear in LLM responses. And it sounds great in theory, but if you look for data on the overlap between the brands and pages that appear in the first page of Google results and the brands and pages that appear in LLM responses for the same query typed as an AI prompt, we see overlaps as low as like 8%. So if you're a top 10 link in Google results for a keyword and then you pop that into chatgpt as a prompt and you only have an 8% chance of showing up, that doesn't sound like good SEO is good geo. And, and the fact is we still don't know why that is and everyone's just rushing to try to figure that out.

[C]

Yeah, yeah, no, it's a fascinating topic. Let's switch up a little bit. I got a question for you, Ari. When you were in San Fran, did you see billboards like this?

[A]

Yes, I did.

[C]

You did?

[A]

Yes. So for those on the audio, it was. So it is the billboard from the company I used to be on the board of called Vibe, where they have a egg like forehead with I would say oblong, Oblong Farhad. And then the headline is Target Mark on tv. So you can show your ads to Mark. Mark is the owner of the Oblong Farhad, whose name should not be mentioned because he blocks me on Twitter.

[C]

So again, this is a fun piece. Also in the New York Times, they had a, a profile of the work that Vibe has been doing and it's super fun. Everybody should, should read it. It's, you know, basically they have these strategically placed billboards all around SF, including along the 101 with messaging like this. Target Mark on TV with an oblong head. Target Elon on TV with the kitchen zinc from the day he bought X. Target Jensen on TV with a leather jacket shoulder. It's, it's really well done and it's, it's, it's. I think this is really good marketing. And again, I give Vibe a lot of credit. It shines a light on how, you know, out of home can be used so strategically because they're just targeting startups to advertise on tv.

[A]

There, there was this whole case study. I think it was Brex. I'm trying to remember. It was a finserve company. I think it was Brex, where they effect got most of their momentum by, by blanketing San Francisco with out of home ads. Bank for startups, right?

[C]

Yeah. Let me, let me give you guys an example. You'll probably remember probably 20 years ago, buddy Media was buying all of the billboards in the airports like Newark, LaGuardia, JFK just, you know, targeting agencies and brands that are Going in and out with this messaging of it was like power tools for Facebook. Something super, super simple. But they gave it a ton of credit for their early momentum as well.

[A]

Yeah, that's a good exit.

[C]

Great. One of the original kind of 60, 70 baggers of new York tech.

[A]

Yeah. One of the great things about out of Home is it's it. People think it's not targeted because anyone can walk by it. But neighborhoods and locations have very specific demographics, so it's quite targeted if, if you choose wisely.

[D]

I mean, the, the, the wastage on it is funny to think about though. The people in that neighborhood who have no idea what it is or don't care, probably don't know if that's a forehead or a thumb or something else. I remember maybe 15 years ago, in the relatively early days of Facebook hyper targeting, I decided to target an ad to a friend of mine who lived in Vancouver. And I knew the things he followed and liked on Facebook. I was able to create an ad that was apparently going to reach about 15 or 25 people, and they let me buy it. And I was able to send him an ad basically making fun of him because that's what I was doing. And I just really enjoyed thinking about the couple dozen people who were not Brett, who saw that ad and didn't know what the heck was happening.

[A]

Yeah, the platforms all have restrictions against pinpoint targeting individuals, so you can't upload a. Upload a list of one email, but there's so many ways around it. Do they. I thought they prevented that, but you could get around it by doing. Exactly.

[D]

I didn't do it with this email. I did it with a very specific set of interests that were able to generate a very small audience.

[A]

So we have Arthur from Vibe coming to market live. And I don't know what the topic is yet of what he's talking about, but I'm trying to convince him to come out dressed like Sydney Sweeney. Like that would be.

[C]

Be.

[A]

That would be the obvious thing for those. This is a little bit of a callback to Vibe's last stunt where Arthur did the American Eagle jeans commercial beat for beat. So we'll see what we can convince Arthur to do.

[C]

He did in the, in the Times piece there. There was some interviewing of people on the street, just kind of like normies. Like, what do you think of these ads? Do you understand these ads? And I think, you know, your instinct that people just don't understand it is right on, Nate. But again, the CAC and the profitability of getting a Startup that is going to start spending significantly, I think more than pays for it. The article quotes that these billboards on the sell side to buy a placement costs between 15k and low 6 figures for 6 weeks. Makes me think this is a business of just really high value locations. Maybe it's only during a certain time period. Just like buying these things.

[A]

Hey Eric, I have a suggestion for you and Joe on your other podcast, which is how important is it to get buzz in Silicon Valley even if you're not a Silicon Valley company? Because that's really what he's doing here. It's not that he expects to get a lot of customers who are based in San Francisco. It's that he wants people in San Francisco to have heard of Vibe.

[C]

He's in the New York Times. He's got people posting this thing incessantly on X. I'm sure Arthur can point you to a couple posts that probably went viral. I mean it's super important. It's like earned media. Probably pays for it itself. Cool, let's move on. Did you see this one? It's your rebate dot com. I'm guessing you did not.

[A]

I did. It's a great topic. Hit it.

[C]

Yeah.

[D]

Yeah.

[C]

So the indie agency Acadia launched insurerebate.com calling out rebate practices at Holdcos. The CEO, Jared Belsky, he posted a couple things. So first like sort of like an op ed in some industry trades and then a post on LinkedIn saying that he paid 53 clients their share of 2025 rebates and challenged other agencies to do the same. This is a thorny topic. So according to Jared, rebates and non transparent incentives are estimated at between 3 and 10% of the $600 billion media market. Digital media market, which is a lot of money, billions of dollars. Industry benchmarks put average agency operating margins at 8 to 12%. Yet rebate income can account for 10 to 30% of total agency profit. So super important. And he's calling out the practice. I feel like this comes back around every few years and there doesn't seem to be be much on the marketer side that changes to force this. But I applaud Jared for taking a creative approach. And again, similar to Vibe, I think the earned media he's getting for this is probably really good practice for him and his agency.

[A]

Yeah, I mean it's one of those things where logically if you're happy with your agency's results, you shouldn't care about what ingredients they put into into the stew. Right.

[C]

That's every Holcos answer to this.

[A]

Right, right. But you do have to wonder if they're making the right decisions for you. To give one example, I'm familiar with the Google's incentives where they combine your spend on the tools, the Google Analytics, Google campaign manager tools, plus media on YouTube, plus media on AdWords, then you give you a rebate, those sort of things. You have to wonder if you're a client, why am I using this particular tool? Why aren't I on the trade desk? Or why aren't I using Flashdog talking? You know, and they may be steering you, they probably are steering you towards products that, you know, you're not, are not optimal for your situation.

[C]

Yeah, makes sense. We're about at time, but maybe we should touch on this one. Since advertising was probably a big catalyst for this. There is a new number one on the Fortune 100. Amazon is now the Fortune 1. Ending Walmart's 13 year run. So Amazon did 717 billion in revenue last year, beating Walmart by 3 billion. Congrats to Amazon.

[A]

So this is ranked by revenue, the Fortune list. All right. Yeah, congrats to Amazon. I guess the, the question is, you know, how much do they keep and you know.

[C]

Yeah, yeah.

[A]

Where it ends up.

[C]

Sure.

[A]

Could we do a shout out to liftoff for withdrawing their ipo?

[C]

Liftoff, Withdrew their IPO and then confidentially refiled later in the day.

[A]

What does that mean?

[C]

So they withdrew their IPO and then they filed their IPO later on that day. It means there was something that happened in a six to eight hour period that made everybody just say like let's forget this thing, this could be really bad and step on the gas. There's been no reporting on this outside of the fact that it happened. So I think we got an ad tech IPO coming.

[A]

Is this like, do you do a diff between the S1 before and the S1 after and see what changed? Is that the, is that the idea here?

[D]

I mean there are things you can and can't do when you're actively filed. Right. Is there something they needed to take care of somebody?

[C]

Everybody went, oh, it's the SaaS apocalypse. Like, you know, it's an example of a company pulling. But I think Nate, you're probably right in that, you know, there might have been something in the filing that needed to get changed. I don't know. So watch the space or watch us space. Figure it out.

[A]

Yeah. Looking forward to the new ipo.

[D]

Yeah.

[A]

All right, well, well, that does it today. Nate Elliot, thank you so much for coming on. Where can we find you? I understand. Understand you have a podcast or eMarketer has a podcast. What's the deal?

[D]

Yeah, eMarketer has a few podcasts. The one you can find me on on a regular basis is called eMarketer behind the Numbers. And of course, eMarketer.com is where I publish research and we have lots of articles, forecasts, surveys and other fun stuff there.

[A]

Wonderful. Well, thanks for being here. This was great conversation. Eric, always a pleasure. And we will see you at marketecture Live in a couple weeks.

[C]

See you next week, everybody.

[D]

Thank you.

[C]

Thank you for subscribing to marketecture.

[D]

New interviews are added every week at marketecture tv.

[A]

And your favorite podcasting.

[C]

APPLAUSE.