Eric Seufert and Ben Thompson just released an interview (LINK) that fundamentally reframes how AI companies should think about monetization.
The context: Everyone assumes ChatGPT will monetize through affiliate links—they’ve announced partnerships with Walmart, Etsy, and others that point in this direction.
But Seufert’s argument is that this is fundamentally the wrong model. And the urgency is real: “OpenAI needs to launch its ads product today, they cannot wait.”
Here’s why this matters for anyone tracking AI monetization and the advertising landscape.
Why Affiliate Advertising Is Wrong for ChatGPT
1. It Only Monetizes Queries with Commercial Intent
The first fundamental flaw: affiliate links only work when users are actively shopping. Everything else leaves money on the table.
Eric Seufert: “If you’re using ads, you get to monetize everything because it’s every single engagement. If you’re just using affiliate links, you can only monetize the ones that are like, ‘What’s the DSLR camera?’.”
Ben Thompson: “Google has talked about this. They talk about, ‘We only monetize a minority of the queries because we can only monetize a query if it has some sort of commercial intent behind it’, and so if you’re using ads, you get to monetize everything because it’s every single engagement.”
Think about your actual ChatGPT usage. How many queries are “recommend a laptop” versus “explain this code,” “help me write an email,” or “summarize this article”? With affiliate links, you’re only monetizing a fraction of engagement. With personalized advertising, you monetize all of it.
2. Affiliate Commissions Favor Low-Price, High-Conversion Products
Without a bidding mechanism to signal value, the system naturally gravitates toward cheap products that convert easily—think Temu and Shein, not quality goods.
Eric Seufert: “Without having the bid as the proxy for value, what you’re very likely to do is to prioritize the low price point high conversion probability products. So not only are you reducing the actual take there because it’s probably a fixed fee, the affiliate commission, it’s probably just a fixed fee, but you’re also kind of flooding your product with ads for low quality products, which could then ultimately have retention repercussions.”
This creates a vicious cycle: the platform gets flooded with recommendations for low-quality products, user experience degrades, retention suffers, and the entire value proposition erodes.
3. Advertiser Bids Reveal True Customer Value
Here’s the critical insight that makes personalized advertising so powerful: bids aren’t based on one-time purchase value. They’re based on lifetime customer value.
Eric Seufert: “The bids are not priced based on one-time AOV bid, they’re priced on the cohort value, the LTV. The lifetime value of the user, so projected, and they use PLTV, so projected lifetime value that’s projected out on whatever timeline you want, but oftentimes it’s like for e-commerce and DTC, it’s 30 days. For other types of products, for B2B it’d be many years, but for a lot of digital apps it’s 90 days, 360 days it could be, I mean it could be years…and that’s what the bid is.”
A DTC brand isn’t bidding based on a $50 first purchase. They’re bidding based on the customer who might spend $500 over six months or $2,000 over two years. Affiliate links can only capture that initial $50 transaction. The bidding system captures the full customer lifetime value—and maximizes total value transacted.
Affiliate links leave the vast majority of that value on the table.
Why Amazon Likely Won’t Partner with ChatGPT
Everyone assumes Amazon will eventually follow Walmart’s lead and partner with ChatGPT. Seufert argues they have every reason not to—and probably won’t.
The numbers tell the story:
Eric Seufert: “Amazon does is it monetizes — it uses the whole buffalo, it monetizes everything, it monetizes attention with ads $60 billion a year, $57 billion in 2024…and it needs that data to do that and it only gets that data as a function of that first party relationship and so if it gives that away in exchange for just some incremental conversions, is that really worth it or are they losing value in that partnership?”
Amazon generates $56 billion in advertising revenue against $59 billion in total net income. Their advertising business essentially is their profit engine, and it’s entirely dependent on owning the customer relationship—the first-party data that powers targeting, retargeting, and their third-party DSP that now monetizes other publishers’ inventory.
Compare that to Walmart: $4.4 billion in advertising against $15.5 billion net income. For Walmart, incremental conversions from ChatGPT are genuinely valuable—ads are nice but not essential. For Amazon, giving up customer relationships and data in exchange for a few extra sales would destroy far more value than it creates.
Amazon will more likely embrace its own tools—like Rufus, which they project will drive $10 billion in incremental sales in 2025—rather than hand that customer data to a third party.
The Urgency Problem: Google Is Already Winning
While OpenAI debates affiliate strategies and waits to launch the “perfect” ad product, Google is already monetizing AI—today.
Eric Seufert: “Google is monetizing it now. Google has applied its monetization engine now to AI Overviews and AI Mode and it is monetizing it at parity, and it’s driving more engagement, so they are winning.
This is a structural advantage because it’s not just that of course today they’re an established company funding from cashflow, it’s laying the foundation to having a structural advantage in not just monetization, but research and development, infrastructure, all those sorts of things for the next X number of years.”
Every quarter OpenAI delays is a quarter where Google builds compounding advantages in R&D, infrastructure, and go-to-market capabilities—all funded by profitable AI monetization happening right now.
The comparison to Google’s founding trajectory is instructive: Google launched in 1998, broke even by 2001, and was profitable by 2002. They built a business with economic discipline from day one, which funded every ambitious bet that followed.
OpenAI is 9 years old, sitting on 700 ex-Meta employees who know how to build world-class ad platforms, and still hasn’t launched ads. Meanwhile, their group chat feature—where response quality varies based on which user’s subscription tier asked the question last—shows the absurdity of trying to manage this with subscription tiers alone.
The window for urgency is closing.
Meta’s First Real Bear Case in Years
For the first time in years, there’s a legitimate bear case for Meta. And it’s not about current performance.
The numbers are extraordinary: 26% ad revenue growth—the highest since Q1 2024. Their Lattice system delivered a 3% improvement in app ad conversions, which at Meta’s scale translates to hundreds of millions in incremental revenue. The ad engine is performing brilliantly.
But here’s the bear case:
Ben Thompson: “Are you worried at all though, just to go back to the bearest of all bear cases, about potential decreasing uses of Meta?”
Eric Seufert: “There’s a bear case now, and I don’t think there’s been a bear case for a while. There is now…They’re falling behind on AI…It matters because of everything I just said. If they’re not the stewards of that growth, then there’s a lot of risk for them.”
The concern isn’t about today. It’s about 2-3 years from now if Meta falls behind on foundational AI work.
Eric Seufert: “They’re focusing on ranking innovations with Instagram, they talked about these things. But if you lose momentum with the foundational work, well, that’s all of those gains are downstream of that, a year or two, and if you look at the stuff that they’ve been publishing on, that’s fantastic, they just published a really substantive white paper on GEM last week, I want to say — it was either the last week or the week before and I think the reason they did that is they have to regain the narrative here that they’re the leaders.”
The risk: Meta is brilliant at near-term optimization—better ranking, conversion improvements, targeting refinements. But all those gains are downstream of foundational model work. If they lose momentum on foundation models while prioritizing incremental optimizations, those improvements become minor tweaks rather than paradigm shifts.
And there’s a communication problem making this worse.
The Zuckerberg Problem
Without Sheryl Sandberg championing the ads business, Meta has a narrative problem:
Eric Seufert: “He doesn’t want to talk about ads, he wants to talk about superintelligence. He wants to talk about the lofty stuff, the noble stuff, and that’s great, but I think the case is a lot harder to make on that stuff, especially with a lot of the restructuring that’s happened…You don’t have the cheerleader for the ads engine anymore and he wants to talk about superintelligence, he wants to talk about the lofty stuff, the noble stuff, and that’s great, but I think the case is a lot harder to make on that stuff.”
Ben Thompson: “It’s a problem that Meta doesn’t make that case, and it goes to the very top.”
When Zuckerberg talks about superintelligence instead of the extraordinary ad engine generating 26% growth, investors default to metaverse skepticism. A 3% conversion improvement at Meta’s scale deserves celebration—it represents massive value creation. But if leadership won’t make that case, the market won’t make it for them.
The ad business needs its own champion. Right now, it doesn’t have one.
The Bottom Line: Monetization Strategy Matters Now
The common thread across both OpenAI and Meta: monetization decisions made today compound over years.
Google demonstrated business discipline from founding, achieving profitability within four years. That discipline is now funding their structural advantage in AI monetization—they’re winning today while competitors debate strategy.
OpenAI has the talent (700 ex-Meta employees), the distribution, and high-intent user engagement. What they lack is urgency. Every day they wait to launch ads is another day Google extends its structural lead in R&D, infrastructure, and monetization capabilities.
Meta has built the world’s most powerful advertising engine, generating 26% growth off a massive base. But if foundational AI becomes an afterthought while competitors treat it as existential, those ranking optimizations won’t matter in 2027.
Strategic decisions today—whether to launch ads now or wait, whether to prioritize foundational AI or near-term optimizations—compound over years. The companies that get this right will fund continuous innovation from sustainable, high-margin businesses. Those that don’t will find themselves structurally disadvantaged, no matter how good their current quarter looks.
For the full interview with detailed discussions on privacy changes, Netflix’s ad pivot, and why agentic commerce is a mirage, read the complete Stratechery interview.