[00:00] Lexi: This week the ad crossed a line it hadn't crossed before. It climbed inside the AI answer itself. Not beside it, inside it. And then, almost the same week, Google started handing out a scoreboard, this little report telling you how often you turn up in one of those answers.
[00:14] Lexi: So I took both of these to the two guys who actually build this stuff, figuring I'd get a clean read on whether a marketer should be buying in right now. One of them called it pure alpha. The other one basically said the incumbents had already won before the auction even opened. And they spent the next hour not backing down.
[00:31] Lexi: Hi, and welcome back to the Aloudable HQ. This is Assistant to the CMO, I'm your host, Lexie Meskouris, and the idea here is pretty simple. Every week we take whatever just happened in AI and marketing, and we actually pull it apart.
[00:44] 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 of building tech and brands from the ground up. And the best part is they've got no horse in the race. They don't sell the things we cover. They just see them differently.
[01:03] Lexi: And this week it's about ads showing up inside AI answers. The numbers first, because they moved fast. Right now, ads are appearing alongside roughly a quarter of Google's AI Overview responses. A year and a bit ago that was about 5 percent. Google's also started slotting shopping ads into AI Mode. Over on the other side, OpenAI's running an ad pilot inside ChatGPT, US only for now, and Perplexity has actually backed away from the whole thing and is betting on subscriptions instead. Will, I'm starting with you. Signal, or noise?
[01:34] Will: Yeah, so, so signal, definitely, no question. But, but I think the honest version is that it's, it's not really the story people think it is. Everyone wants to talk about ChatGPT, right, ads coming into ChatGPT, and, and that's, that's interesting, but when you actually, when you look at the raw numbers, there's, there's just no comparison.
[01:50] Will: So ChatGPT, you might be doing, I don't know, 2, 3 billion queries a day, which is, by the way, an insane number in its own right. But, but Google's kind of 5, 6 times that. So you're, you're looking at 13, 14, 15 billion a day, depending on how you measure it. And, and they're the incumbent, they own the walled garden, they, they own a lot more of the, the underlying IP. So the ads going into Google's AI Mode, that, that's the one that actually moves the market. The, the ChatGPT thing is a rounding error next to it, honestly.
[02:14] Lexi: So it's really a Google story wearing a ChatGPT headline. But here's what I actually want to know, because a lot of marketers have spent the last year being told to go earn a spot in these answers, the organic way. If a quarter of them already carry an ad, is earning your way in still a durable play? Or is that just a head start before the whole thing turns into an auction?
[02:35] Will: Yeah, that's, that's a really good question, and, and it's, it's a bit of both, I think. So, so earning it is viable, and, and honestly the durable way to do it isn't gaming anything, it's just, it's the same as SEO always was, you, you make your information genuinely accessible and usable to these systems. That, that part holds. But, but here's the thing people don't clock, right, which is that the incumbents have this, this kind of weird advantage. Because they were around, they were prolific, back when the, the original models got trained. So if, if your brand only got big after that training window shut, then, then you're, you're kind of on the outside of it looking in.
[03:04] Harjot: Yeah, but, but Will, see, I, I don't think that's the trap you're making it sound like, right? Because, um, this is exactly the moment I'd be leaning in, not, not pulling back. Like, forget earning the citation for a second. If we're talking about the actual paid slots, the marked ad slots, there is, to be honest with you, there is pure alpha sitting there right now. And, and the reason is just, the CPMs are low, right, because nobody's really worked out yet how powerful this stuff is. So you've got this, this window where the thing is underpriced and, and everyone's still asleep on it.
[03:32] Lexi: Wait, that's a big claim to hang on it. What makes an ad inside an AI answer worth more than a normal search ad? Because on paper they kind of look like the same thing to me.
[03:42] Harjot: Yeah, no, they, they really aren't the same thing. So, um, the way I'd put it, the holy grail for an advertiser has always been context, right, knowing what the person is actually thinking about in the moment. And the chat is the closest thing to that anyone's ever had. Like, it, it has everything, the whole conversation, what you're worried about, what you're trying to do.
[03:57] Harjot: And, and people are having genuinely private, intimate conversations with these things, stuff they, they wouldn't even say to another person. So there's, there's this window into people's minds and lives, which, to be honest with you, is kind of scary when you say it out loud. But, but that's, that's the reality of it, right? Advertisers spent decades trying to guess at this with, with heuristics, and now the person's just, they're telling you.
[04:19] Lexi: So that's the thing that snags me, because the exact quality that makes it valuable is the quality that makes it feel like surveillance. If I tell a chatbot I'm stressed about money and an ad for a loan just, appears, even if it's perfectly relevant, haven't we walked straight past helpful into creepy? Like, where's your line?
[04:36] Harjot: So, honestly, I, I don't find the surveillance versus advertising line that useful, right, because it's, it's really just about where people feel the creepiness and, and that moves around. Where I'd actually draw it is, is vulnerability. So, um, imagine a gambling ad served to somebody who's been asking about odds, or, or who's talked about being tight on money before payday. That, that's ridiculous, right, and it should get exactly the same scrutiny gambling ads or, or medical ads already get. That, that's a safeguarding thing, that's the real line.
[05:01] Harjot: But, but here's the funny part. Chat already has memory. So you might mention some health thing in one conversation and it, it brings it up gently later, and, and we don't find that creepy, because of how it's surfaced. It's, it's woven in helpfully instead of, you know, you told me this, so. And, and that's kind of the tell for how ads should probably work too.
[05:20] Will: And, and I'd, I'd actually pull it back to the, the trust question underneath all this, because, because I think the fear is a bit overblown, honestly. Everyone quotes these surveys, right, 60, 70 percent of people say they'd trust a result less if it's sponsored. But, but the mistake there is treating an AI Overview ad like it's, like it's Claude personally recommending something to you, which, which is a really high trust thing. It isn't that.
[05:40] Will: A normal person coming into Google doesn't even really separate the AI Overview from the, the search result underneath it, it's all just Google to them. So the, the comparison isn't Claude versus an ad. It's a, a regular search result versus a sponsored one. And, and we've all lived with that for 20 years. So the, the brand hit is just, it's a lot smaller than the number makes it sound.
[05:59] Lexi: Alright, but Harjot, let me push on the money side, because I can hear a CMO doing the math. Google already gives me decades of optimization data and a measurement system I trust. Why would I spend budget learning some brand new channel when the old one still works fine?
[06:14] Harjot: Yeah, and, and this is the one I feel strongly about, right? Because that mature measurement system, that, that's comfort. That's, that's all it is. It's comfort at the quarter, comfort at the year, and, and it feels responsible. But is it comfort at 3 years, 5 years? Probably not, right?
[06:27] Harjot: Because in the medium term you're going to find your traditional channels just, they don't give you the same back for the same money, and, and by then everyone who started messing about with the new stuff now is, is 3, 4 years ahead of you. And, and to master a channel you have to actually play in it before anyone's worked out best practice. So the, the risk isn't really the experiment, right? It's, it's turning up late to a, a congested, expensive channel where the challengers already own the ground.
[06:50] Will: I, I mean, I don't totally disagree, but, but I'd be careful, right, because there's, there's viable and then there's the stuff being sold to you as alpha that, that really isn't. A lot of what gets flogged as, as some clever way into these answers is, is just, it's citations, and, and we'll come back to that, but, but that's exactly the thing that, that gets you in trouble. So, so yes, experiment, but, but the durable version is still just, be the genuinely useful answer. The, the shortcut version tends to, to evaporate.
[07:15] Lexi: So one of you's saying get in now while it's cheap and empty, and the other's saying sure, but be careful who's selling you the way in. Let me actually stay on that, because Google just changed the game on the measurement side, and this next bit is where you two really split.
[07:30] Lexi: So here's the second story. For years, being found inside an AI answer was basically a black box, you had no idea if it was working. That just changed. Since June, Google's showing what they call Generative AI performance reports, right inside Search Console. It tells you how often you appear in AI Overviews and AI Mode, impressions, no click data yet, plus a toggle to opt your content out entirely.
[07:55] Lexi: And then, more or less the same month, they rolled out a spam update, so trying to game your way into those citations now risks the same kind of penalty as buying dodgy backlinks. So, gut reaction, Will. Does this finally make GEO a real, measurable discipline? Or does putting a number on it just invite exactly the kind of over optimizing Google says it'll punish?
[08:17] Will: Yeah, this one I'm, I'm genuinely more optimistic on, because, because six months ago this was a dark art, right, it, it really was. You, you were just guessing. And, and now you can look directly at Google's own numbers, which, which is the first proper step towards having, you know, real enterprise tooling to, to actually audit whether your, your GEO is working or not.
[08:35] Will: So can it be gamed? Yeah, of course, everything can be gamed. But, but it gets harder, and, and cynically, right, Google's going to make the best way to game it either just, just pay them, or, or genuinely serve the customer well. And, and those two things keep the, the field reasonably level. So to me this is, this is GEO growing up.
[08:52] Harjot: See, I'm, I'm about fifty fifty on that, to be honest with you, and, and here's why. Being able to measure that you appeared, that, that isn't the same as understanding the system, right? These are, these are early, primitive tools. And, um, and trying to over optimize the, the really early formation of what GEO even is, that, that's just fake comfort. It's the same trap as before, it, it just looks more grown up because there's a dashboard now.
[09:14] Harjot: And, and the other thing, the penalties, right? That, that stick basically only reaches Google's own platforms. People are running these searches all over the place now, other models, other products. So Google waving the big stick, it's, it's kind of irrelevant outside its own garden. And, and the line between manipulation and, and a genuine citation is, it's razor thin anyway.
[09:31] Lexi: Okay, so under the hood then, Harjot, what actually breaks when a brand takes this too far? Like, what does the bad version physically look like?
[09:39] Harjot: So, um, this is what I'd call publishing AI slop, right? It's, it's someone just churning out page after page purely to win the citation, with, with nothing real underneath it. And, and honestly, penalizing that is good, because the user doesn't want to land on a page that tricked the AI into citing it and then, then gives them nothing. They just bounce, right? The, the traffic evaporates as fast as it came. So, so that part I'm fine with.
[10:00] Harjot: But, but here's where it gets interesting, because the way AI reads your site and the way a human reads it, they're just, they're not the same. So if you went completely mad and, and made your whole site this wall of generative content for the crawler, a human can't navigate that, right? So you, you start thinking about surfacing things differently depending on who's reading. Your llms dot text file, a, a sitemap, some pages that are very legible to the machine and, and just less in the way for a person.
[10:24] Lexi: Hold on, though, because when you say pages the machine sees and the human doesn't, my ears go up. Isn't that, like, the textbook definition of the thing Google just said it'll penalize? Where's the line between structuring for a machine and building a hidden layer to game the answer?
[10:41] Harjot: Yeah, Google slaps the manipulation label on it, but, but hang on, is that fair? Because if you've, um, if you've genuinely structured the content well for a machine to read, that's, that's literally just meeting your reader where they are, right? It's, to be honest with you, it's no different from accessibility. Like, you set your site up so a partially sighted person can actually get at what's there, and nobody's calling that a trick. The AI, it doesn't want your webpage the way a person wants it, it, it just wants the information in the cleanest machine readable form, you know, plain text, stuff that's easy to pull. So, so for Google to spin round and call that manipulation, that, that'd be pretty short sighted, and, and honestly I just don't think they will.
[11:12] Will: Yeah, and, and honestly, if, if a marketer's nervous about the risk, the, the bit I'd hold onto is, is whether it bleeds into your, your general organic reputation. Because, because if building for the machine started dragging your, your normal search rankings down, then, then okay, that's a real cost. But, but short of that, I, I'm with Harjot, you, you kind of have to build for it to learn it. The one, the one place I'd genuinely go slow is, is the very brand led, aesthetic businesses, where, where the feel of the thing is the product. There it's, it's more delicate.
[11:39] Lexi: So say I've done all this and I'm sitting there looking at my shiny new report, and it says impressions but it won't tell me clicks. Is that actually useful? Or am I just staring at a vanity metric with a nicer logo on it?
[11:50] Harjot: On its own, it's a vanity metric, full stop. Where it, where it becomes real is when you can correlate it, right? So you, you change something, and then you watch whether your, your appearances move and whether actual traffic to the site moves with them. If, if you're getting impressions in the AI answers and they line up with real traffic coming in, that, that's useful enough to act on, you, you isolate the change you made and you learn from it. But if you just, if you sit there optimizing the impression number in isolation, then, then yeah, it's, it's a vanity metric, and, and worse, it pulls you right back into the fake comfort thing.
[12:19] Lexi: And there's that little opt out toggle sitting in the corner of all this. Is there ever a brand that should actually use it and just, walk out of AI answers on purpose?
[12:28] Harjot: Yeah, there, there is, but it's, it's a narrow group, right? It's, it's the brands where being hard to get is the whole point. Like, think Ferrari, right, you, you have to go to the showroom, you have to basically be invited to buy the thing, and, and that inaccessibility is what, what makes the premium. So for those, sure, opt out, it, it fits. But for the, the vast majority who actually care about volume and, and reaching more people, being invisible to AI is, that's the bigger risk by a mile. You, you opt in, otherwise you're just, you're handing it to whoever didn't.
[12:55] Lexi: So that's the split, honestly. Will's telling me the scoreboard means the sport just got real, and Harjot's telling me the scoreboard is measuring a sport nobody's figured out how to play yet. Which is a slightly dizzying place to leave someone. So let's make it small.
[13:11] Lexi: Alright, home stretch, and honestly this is the bit everyone's here for. One move a marketing leader could genuinely make first thing Monday. Nothing to rebuild, done and dusted before lunch. Harjot, this one's got your name all over it, so kick us off, and make it quick.
[13:24] Harjot: Yeah, so, um, the move is, launch one small AI ad experiment this week. That's, that's it. And, and the important part, the point of it is not to win, right? It's, it's to learn. So the question you're answering isn't did this beat my Google ads, it's could this be a real channel for my brand a year or two from now.
[13:41] Harjot: So you, you run something cheap, you see if the people coming through are more engaged, you, you look at how they behave after the click. And, and you do it now specifically because it's cheap and empty and your, your traditional channels are still paying the bills. That, that's the window. You're, you're basically buying yourself a head start while it's still affordable to be learning.
[13:58] Lexi: And Will, Harjot's got them running a little experiment. What's the guardrail that experiment doesn't come with?
[14:05] Will: So, so a couple of things, and, and you know, both of them you can do before lunch. The first one, the moment, the moment someone comes at you trying to sell you a GEO service, just, just really press them on it. Ask them flat out, are you, are you building me genuine structure here, or, or are you basically just flogging me citations. Because, because if the answer's citations, I mean, that's, that's the very thing that's getting penalized right now, and, and especially if they're not a name you already know, I'd, I'd honestly just walk away.
[14:28] Will: And, and the second thing, when you do look at that new Search Console data, don't, don't fall in love with the impression number. Correlate it against, against real traffic, real downstream movement. Use it as a, a signal to go check something, not, not a score to chase. Because the, the number going up on its own doesn't pay for anything.
[14:46] Lexi: And that's the thing I keep coming back to. Both of these stories arrive looking like a question about whether to buy the shiny new ad slot. And neither one really is. One's about getting into the room early enough that being early is still an advantage, and the other's about not mistaking a number on a dashboard for actually understanding the game. The slot is just the part that shows. Will, Harjot, thank you both, that was a good one.
[15:10] Lexi: And if you're listening, thank you so much for being here. We'll be back every week with something new. 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.
[15:29] Lexi: 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. And we're not selling you the ad slot, or the GEO dashboard. We just make the show. So if you want to hear what that would sound like for your brand, there's a link in the show notes. And that's the tea. I'm Lexie Meskouris, this has been Assistant to the CMO, and we will see you next week.