[00:00] Lexi: This week, OpenAI started putting ads directly inside ChatGPT answers. And at the same time, a bunch of brands are loudly swearing off AI entirely. So today we're getting into both sides, what the ads actually look like, why companies are slamming that door, and whether any of it holds up.
[00:18] Lexi: Okay. So. Hey, everyone, and welcome from the Aloudable HQ. This is the first ever episode of Assistant to the CMO, and I'm your host, Lexie Meskouris. And here's the idea of the show. Every week, we take the hot button stuff in AI and marketing, and we actually break it down.
[00:34] Lexi: Now, I don't build any of this myself. So each week I sit down with two people who do. Will Nash and Harjot Singh. Between them it's, honestly, over 20 years in tech and building brands from the ground up. And the nice thing is, they've got no horse in any of it. They don't sell the stuff we cover. They just see it differently.
[00:52] Lexi: So every episode, we'll get into the latest in AI and marketing. And we'll argue about whether each one's a real shift, or just another thing that's about to blow over. And honestly, I'm really glad you're here for the first one.
[01:05] Lexi: So today we've got two big stories. First up, OpenAI quietly putting ads right inside ChatGPT's answers, which is already heading to the UK. And then, the flip side, this whole wave of brands doing the exact opposite, loudly swearing they do NOT use AI. So let's jump right in. Will, I want to start with you. This ads thing. Is it enormous, or is it nothing?
[01:26] Will: Yeah. So, before I answer that, I think it's worth uh picturing it properly, because the picture really matters here.
[01:33] Lexi: Right, because when I hear "ads inside ChatGPT," the obvious version is a banner. Some flashing box off to the side everyone's trained themselves to ignore.
[01:43] Will: Interestingly, It's not actually a banner and it's not off to the side.
[01:47] Lexi: So walk me through it. I ask ChatGPT something, completely normal. What do I actually see?
[01:53] Will: So... you ask it something, you get your answer, the normal answer, and then below that there's a card. A sponsored card. It's labelled. It's got an advertiser name, a little logo, a headline, an image, a link. And it sits right at the bottom of the answer. So it's inside the conversation, not next to it.
[02:14] Lexi: Okay, and here's the bit I keep snagging on. OpenAI is very insistent the ads don't influence the answer itself. So is this a thing sitting next to the answer, or is it inside it?
[02:23] Will: So You're absolutely right to pick up on that. So they're really insistent the ads don't influence the answer. They run on a separate system, advertisers can't touch how ChatGPT actually responds to you. So in their framing, the answer is the answer, and then the ad comes after. But I mean... it's still sitting inside the thing you trust. So it reads like a recommendation. That's the bit I keep coming back to.
[02:45] Lexi: Right, and that's the whole tension, isn't it. You saying - "It sits at the bottom, it doesn't touch the answer" - it sounds clean. But it's still riding on the trust of the thing above it. Okay. So now I want the actual verdict. Harjot, Will's clearly got one eyebrow up about this. You don't. You think this is genuinely big. So tell me, for a UK marketing leader hearing this, what's the thing you'd actually be excited about here?
[03:08] Harjot: Yeah, no, I mean, okay so this is the bit I genuinely get excited about, right? Will's worried about the trust thing, and fair, we'll get to it. But step back. What is an ad, really? It's a guess about what you want. And every ad system we've ever built is basically a worse and worse proxy for that guess. Keywords. Cookies. "People like you also bought." It's all archaeology. We're digging around for clues about intent. But in ChatGPT? People just... tell it.
[03:31] Will: Yeah, they type the actual thing.
[03:33] Harjot: That's it, they type the actual thing! "I'm moving house in August, I've got a dog, budget's tight" — that's not a keyword, you know, that's the real thing. Nobody types that into Google. So to be honest with you, that's the richest read on what a person actually wants that's ever existed. That's the alpha.
[03:47] Harjot: And then, and this is the second bit, it's trusted. You ask it like you'd ask a mate who knows stuff.
[03:52] Lexi: Mm. And the top result in search has been a money game for years now.
[03:57] Harjot: Right, everyone knows the top result's paid. But the model? People haven't built up that armour yet. So you've got intent and trust in the same place. That's new. That's super, super interesting if you're a brand nobody's heard of.
[04:07] Will: Yeah, and that's the strongest version of it, I'll give you that. Intent plus trust in one place. It's a real thing.
[04:12] Will: But here's where I get cautious. The intent's only as good as what they let you do with it. And right now? You can't do much. You can't bring your own customer list. No lookalikes. No retargeting. No demographic breakdowns. You get context matching on the chat, and a little card. That's it.
[04:29] Harjot: That's it, yeah, that's real.
[04:31] Will: So someone tells ChatGPT the whole story of their life, and as an advertiser you're basically catching a wave you can't actually steer. And then the second thing. Measurement. Everything comes back aggregated. Views, clicks, that's it. No idea which conversations triggered you. So you're flying half blind, at a cost that, at launch, was sitting well above Meta. I mean, it's come down a lot, fair. But you've got the richest signal ever, and the thinnest controls we've had in years. That's the bit that doesn't add up for me yet.
[04:59] Harjot: Yeah, no, and I'm not gonna pretend that's wrong, right? The controls are thin. Genuinely. You can't bring your list, you can't retarget, the measurement's basically a clicks and views shrug. That's real. But here's where I think you and me are looking at different clocks. You're holding it up against Meta, against Google, the machine we've spent 15 years polishing, and against that ruler, yeah, it's a toddler.
[05:17] Will: I won't deny that it's early days still.
[05:21] Harjot: But I'm going... it's the second month. The minimum's already gone from 200 grand to nothing. The pricing's collapsed. They're shipping a pixel, a conversions API. That's not a finished thing, that's a thing being built at speed, in public. So measuring this with the old ruler, the attribution, the lookalikes, I think we're sort of judging a new world by the old world's scorecard. The scorecard itself is gonna change.
[05:42] Harjot: And the thing I keep coming back to, it's coming to the UK either way. Mid year. So I'd rather be early, fumbling with thin controls on the richest signal anyone's ever had, than wait till it's mature and the high-quality attention's already been hoovered up. That's the bet. Early and a bit blind beats late and comfortable.
[05:58] Will: Yeah. And I I genuinely respect the conviction there. "Early and a bit blind." I mean, that's that's kind of the whole bull case in 5 words, isn't it.
[06:08] Harjot: Exactly, one hundred percent.
[06:10] Will: But here's where I'd push back. Being early only pays if you learn something the latecomers can't. And right now? I'm not sure you can.
[06:18] Harjot: But why not, though? Like what's stopping you from learning?
[06:21] Will: Because you can't keep an audience. You can't build a list off it. The signal's the richest ever, sure, but it evaporates the second the chat ends. So the question I keep asking is, what's the durable advantage of being in early? On Google, on Meta, early movers compounded. They built data, they built audiences. Here, you're renting attention by the click and learning nothing you get to keep. So I'm not saying don't test. I'd test. I just wouldn't move budget. I'd put a little money in, watch what actually converts, and keep my powder dry until the scorecard you're betting on actually shows up.
[06:52] Lexi: Okay, hold on, can I just pause you both for one second. Because you keep using a word and I want to make sure everyone's with us. Attribution. When you say attribution, you mean knowing which ad actually caused the sale, right.
[07:05] Will: Right. So, you know, did this person buy because they saw my card, or were they were they always going to buy anyway.
[07:13] Lexi: And that's the piece that's completely missing right now.
[07:16] Will: That's that's exactly the thing that's missing. You just you just can't see it yet.
[07:20] Lexi: Okay. Good. So I get the business case, both sides of it. But Will, you said something earlier and then we kind of skated past it. You said it reads like a recommendation. And that feels like a bigger thing than thin controls. Because thin controls is a "this isn't worth my money yet" problem. But "it reads like a recommendation" is a "should this exist" problem. So say the quiet part. What are you actually worried about?
[07:42] Will: Yeah. So. Thin controls is a money problem. It'll sort itself out, you're right about that. But this is a different thing. So here's the worry. Old advertising, the stuff we've spent our careers on, it mostly catches demand that already exists. Someone wants a thing, they go looking, you put your hand up. You're capturing intent.
[08:02] Will: But inside a chat? You can generate it. Someone's telling the model the whole story, they're open, they're persuadable, and the answer they trust hands them a recommendation in the same breath. That's not capturing demand. That's creating it. And I mean, OpenAI says the ad doesn't influence the answer, and fine, technically, separate system, I believe them. But it doesn't have to touch the answer. It just has to sit underneath it, wearing the answer's clothes.
[08:27] Will: So the honest version of what's happening isn't attribution. It's persuasion. And that's a slightly troubling way to make money off someone at their most open. That's the bit I can't shake.
[08:37] Harjot: Yeah. No, and, that's the strongest thing you've said, right, and I'm not gonna wave it away. "Wearing the answer's clothes." That's a real worry.
[08:44] Lexi: And yet somehow you don't end up in the same place.
[08:47] Harjot: I land slightly different. So OpenAI's saying the ad doesn't influence the answer, and you go "fine, technically," but I think that line's gonna get tested hard, and honestly that's the bit that matters more than the persuasion framing. Because look, every channel we've ever loved was persuasion dressed as something else, right? The friendly voiceover. The influencer who "just genuinely loves this product." You know. The native ad that looks like the article. We didn't invent manipulation in 2026.
[09:11] Will: No, but I think the the intimacy here is new. That's that's a different thing.
[09:17] Harjot: That's fair. That's the the actual thing, right, someone's told the model the real story, and um a recommendation lands in that same warm moment. That's powerful and yeah, a bit uncomfortable. But here's my thing. You can call it persuasion, I can call it the richest intent signal ever, and we're we're both sort of right, you know, because the old words, attribution, persuasion, they were basically built for a world of keywords and banners. We're measuring a new thing with the old ruler. And the ruler bends.
[09:41] Harjot: So my honest read? It's coming. Mid year, UK, and either way, regardless of whatever you and I decide on this mic.
[09:48] Harjot: And I would rather be early while it's still clearly labelled, still excluded from the sensitive stuff, still got someone like that researcher who quit kicking up a fuss in the New York Times, than show up in 3 years when it's all settled and nobody's watching.
[10:01] Will: And and that's the caveat though, isn't it.
[10:03] Harjot: Yeah, 100%, right? Because, you know, that researcher's not wrong, this is an intimate archive. Um, so I I can see why people want pressure on the privacy line, the regulation, EU, UK, all of that. I get it. But um "rather not exist"? Nah I'd rather be in early, with my eyes open.
[10:20] Lexi: Okay. I love that you both ended up in the same place from completely opposite doors. So let me land this with one thing someone can actually do on Monday. Will, you first, and keep it concrete. If I'm a UK marketing leader and this is hitting us mid year. What's the one move?
[10:36] Will: Yeah... So, to keep it concrete, before you spend a single pound on this, find out whether ChatGPT even mentions you.
[10:43] Lexi: And literally nobody thinks to check that.
[10:46] Will: Exactly. And I think the thing that most people don't do, which is kind of extraordinary, is just just ask ChatGPT. You know, turn your memory off so it's not uh flattering you back, and just say, "Recommend me a good X in my category." And if your name doesn't come up, I mean, what what are you doing paying to sit underneath that answer. That's that's just lighting money on fire. So for me that's that's step one.
[11:10] Will: And there's a free tool for it, actually. HubSpot's got an AEO Grader. You feed it your brand, it goes off to ChatGPT, Perplexity, Gemini, and tells you how each one talks about you. Whether you even exist in there. So the one move? Before you buy the ad, check you're in the answer. Earn the mention first. The ad's the second conversation, not the first.
[11:30] Lexi: So, earn the mention first. That's incredibly practical. Okay, Harjot, so you keep saying "be in the room," and I'd love that for everyone, but what does that actually look like for someone who doesn't have a ton of budget to throw at it?
[11:42] Harjot: Yeah, so, you know, Will's spot on there. Um, and mine's basically the cheap version of being in the room. You don't need budget. Honestly you just need to be watching. So find a US startup that's already messing about with this, there's loads of brands public about it now, big ones too, and just... follow them. Read what they ship. Because they're like 2, 3 months ahead of us, and they're learning the new ruler in real time.
[12:03] Harjot: So when it lands here mid year, you're not starting cold. You've watched someone else burn the first 50 quid figuring out what a chat card even is, right? That's the move. Let someone else go first, learn from what they ship, you know, what works, what doesn't, and then you turn up already knowing the playbook. So yeah, check you're in the answer, and watch the early spenders do the hard work for you. Honestly? That's being in the room for free.
[12:23] Lexi: Okay so that's the future OpenAI's building... weave the ads right in, be in the room. But here's what I keep noticing. While they're doing that, a whole pile of brands are running the exact opposite play. They're saying it loud. We do NOT use AI. And I mean loud.
[12:39] Lexi: Aerie just put Pamela Anderson in a campaign where she's literally trying to prompt an AI model, getting more and more annoyed, and the whole thing lands on "you can't prompt this." Real bodies. They've got this decade old pledge, no AI people, ever.
[12:54] Lexi: Then Equinox, "Question Everything But Yourself," real sweaty athletes next to deliberately grotesque AI slop. A steak that turns out to be cake. And the kicker is they made that slop themselves, with AI, on purpose.
[13:10] Lexi: And McDonald's in the Netherlands went the other way, ran an AI Christmas ad, pulled it in about 3 days after everyone called it creepy. 7 weeks of work, gone.
[13:20] Lexi: So. Same week, basically, one company buying its way deeper in, a bunch of others slamming the door. Harjot, you're the AI bull. You said it yourself last bit, you'd rather be early. So when these brands are SLAMMING the door and putting it on a billboard... is that conviction, or is that just the room they picked?
[13:38] Harjot: I mean, is it even one thing? Because they're not the same thing, right?
[13:43] Lexi: And this is what's so fascinating. Equinox made the slop themselves. With AI to dunk on AI. So even the anti AI flex is an AI production now. And then McDonald's, same impulse, opposite result. 7 weeks, gone in 3 days. So actually sort these for me. Where's the line where it's real, and where's the line where it's just the next thing to sell?
[14:04] Harjot: Yeah, okay, let me actually do this and sort them. So, Equinox, I'm completely fine with. Genuinely. They used AI to make the slop. On purpose. The 3 breasted woman, the steak that's a cake. That's not anti AI, that's anti slop. You can only point at bad AI that precisely if you understand the tool, you know? So to me that's the most pro taste thing in the bunch. It's a brand going, the enemy isn't the machine, it's the lack of judgment.
[14:28] Lexi: And McDonald's?
[14:30] Harjot: It's the same with them. And the thing everyone said was, oh, they rushed it.
[14:34] Will: 7 weeks, though.
[14:35] Harjot: 7 weeks, actually. 7 weeks and it was still slop. So time isn't the problem. Right? Time isn't taste. You can spend 7 weeks and have no judgment, and you can spend an afternoon with AI and have great judgment. That's the whole game. The slop is just AI with nobody good in the room. So most of this "we don't use AI" stuff... it's trying too hard. It's putting it on a flag. And a flag is a promise the future can break.
[14:55] Will: Yeah, and that last bit's the one I keep snagging on. A flag is a promise the future can break. That's exactly my worry.
[15:03] Harjot: But you do split off from me, somewhere.
[15:05] Will: I do. Most of them, yeah, it's a bandwagon, they saw the door and they slammed it because slamming the door photographs well right now. But Aerie. Aerie's the one I'd actually defend.
[15:17] Lexi: Wait, why Aerie specifically?
[15:20] Will: Because the reason's dead simple. They didn't start this in 2026. They stopped retouching people in 2014. 12 years ago. So "no AI bodies, ever" isn't a new flag for them, it's just the next line of a story they've been telling for over a decade. It's earned. And I think that's the whole test, isn't it. Not human versus machine. Whether the flag was already true before AI gave them a reason to wave it. Equinox, fine, that's taste, I'm with you. But Aerie's the one that holds up if you pull on it. The others... I'm not sure they survive the tug.
[15:50] Harjot: Yeah, no, and Aerie's the strongest case, you're right, I'll give you that. 12 years. It's true, it's earned, it pulls.
[15:56] Harjot: But here's the bit I still can't get past, even with Aerie. The word "ever." No AI bodies, ever. That's not a value, that's a contract with a future you can't see. And, you know, the thing they actually did well isn't the pledge. It's Pamela Anderson getting more and more wound up trying to prompt the thing. That's the taste. That's the human in the room.
[16:13] Harjot: The "ever" on the flag is the weakest part of the whole ad, to be honest with you. So even on the one you'd defend, I'd keep the craft and lose the promise. Do it so well nobody has to be told. Because the second you have to say it out loud... you're sort of admitting the work isn't doing it on its own. Right? And Aerie's work is. So they don't even need the flag.
[16:29] Lexi: Okay, see, that's a good distinction. Keep the craft, lose the promise. But Will, that's not quite your worry, is it. Because Harjot's saying the "ever" is just weak copy. And I think you're saying it's a trap.
[16:41] Will: I mean, that's not weak copy. That's that's a hostage situation.
[16:47] Lexi: So let's go there. Aerie says no AI bodies, ever. And then Anthropic, totally opposite end, runs a whole Super Bowl ad against ads in AI, and they quietly used Claude to make it. Site visits went up anyway. So is the lesson just don't put it on a flag at all, let the work speak for itself? Or is there a version where you plant the flag and then a year from now you actually need the thing you swore off?
[17:10] Will: Yeah. And that's the one I I actually lose sleep on. And I want to be careful here, because Harjot, you're calling the "ever" weak copy. I I don't think it's weak. I think it's a hostage. And here's the difference. With Aerie, the flag was true before they planted it. You know, 12 years true. So it's not a bet on the future, it's a report on the past. And that holds. But then you look at Anthropic. Same week, near enough. Whole Super Bowl ad, you know, ads are coming to AI but not to Claude, very principled. And they used Claude to make it. And it worked, the numbers went up.
[17:40] Harjot: So that's interesting because, the market DIDN'T punish them.
[17:44] Will: And I think that's that's the bit that unsettles me, honestly, more not less. Because if the gap between what you say and what you do doesn't actually cost you, then the flag was never really the point, the work was. So I I kind of land somewhere close to you, actually. Plant it if it's already true. But the second it's a promise about a year you can't see, you've you've handed someone a stick. And then one day you need the very thing you swore off just to keep up. And then it's not your flag any more. It's theirs.
[18:08] Lexi: And that's the thing, right, one day it's their flag, not yours. I think that's an incredibly good place to land because that's basically the through line of everything we've been talking about today. The shiny thing, the loud thing, the door you slam, none of it's the point. The work is the point. Whether there's taste in the room.
[18:24] Lexi: And I love that we got there from two completely different directions. Will, Harjot, seriously, thank you both. That was an absolutely fantastic first conversation.
[18:34] Lexi: And if you're listening to this, thank you so much for being here for the very first episode of "Assistant to the CMO". We'll be back every week with something new.
[18:42] Lexi: Before we go, I have to tell you something. Every voice you've heard today, all 3 of us, me included, it's all AI. But the opinions are absolutely ours, and the judgment behind every word is human. That's the part you can't replace. And that's what Aloudable does. We take what you'd write, turn it into a show in your own voice, and we keep the taste where it belongs, with a person.
[19:03] Lexi: And we don't sell ads in chatbots, we're not running a billboard about it. We just make the show. So if you want to see what that would sound like for your brand, there's a link in the show notes.
[19:11] Lexi: And that's the tea. I'm Lexie Meskouris, this has been Assistant to the CMO, and we will see you next week.