ChatGPT Ads Hit Europe: Why Your First 3 Months Will Disappoint

Episode Transcript

00:00:00 - 00:00:29

Chris: Welcome to another episode of Growing Ecommerce. Today we’re going to talk about OpenAI launching ads in Europe. Welcome to the least ad platform-friendly market that there is. Have fun! We’re going to talk about how you access this platform, expectations, and how this stacks up currently to Google Ads. All that and more coming this episode. Mike, I missed you.

00:00:29 - 00:00:50

Mike: I missed you too. I’m not kidding. I missed the podcast, and I missed you. I sent you a couple of messages. You were at home—Boston, right? Did you enjoy it? I loved it. Boston is booming. The energy is insane in that city.

00:00:50 - 00:01:11

Chris: I haven’t been back much in the last few years. So you met family?
Mike: Yeah, exactly.
Chris: And you did some business there as well, of course?
Mike: Of course, of course. I started with a conference, and no trip to the U.S. without a business plan! Then I parlayed it into a vacation.
Chris: Good to have you back.

00:01:11 - 00:01:29

Mike: Yeah, good to be back. And your vacation, Chris? All good?
Chris: Yes, good. I did a lot of stuff I actually didn’t want to do, but besides that, all was great.
Mike: That’s the way most of my vacations have been in the past few years!

00:01:29 - 00:01:57

Chris: Man, can I start with something which is important to me?
Mike: Sure.
Chris: All right. We’re talking about ChatGPT again today, right? We have been quite critical of ChatGPT in our last episodes, and I just want to put on Front Street that we are absolutely neutral to ChatGPT.

00:01:57 - 00:02:18

Chris: Actually, I want them to succeed because I think the more pressure there is on Google, the better it is for everyone. While there are not only positives today about ChatGPT, it’s important for me to clarify that we actually want them to succeed.
Mike: Definitely. I think this is very important to listeners here. As always, we spit out the truth and talk about things the way we see them.

00:02:18 - 00:02:51

Mike: My big thing here, Chris, is that I want to see ChatGPT ads succeed, but I like watching Sam Altman suffer.
Chris: And this is the bias on your side!
Mike: I actually don’t like Sam, but my wish for them to succeed outweighs my dislike for Sam.
Chris: You’re a better person than us!
Mike: Of course! But let me jump right into it.
Chris: What’s the news with ChatGPT?

00:02:51 - 00:03:12

Chris: On August 18th—I was on the beach, but I believe it was August 18th—ChatGPT announced that they are rolling out in Europe. About time!
Mike: Exactly, we’ve been waiting.

00:03:12 - 00:03:32

Mike: Europe is last on the list because it’s not an easy market—it’s the worst market, let’s call it what it is, due to laws, rules, regulations, and so forth. That will have a bearing on the platform as well.
Chris: They are live now in 31 new markets.

00:03:32 - 00:04:10

Chris: This is big for them. Demand from our advertisers is super high. People have been asking since it launched, well before it became live here. So there is a very high level of interest, which is positive for OpenAI. It boils down to the fact that people feel so dependent on this monopoly or duopoly of large platforms like Meta and Google, and they’re looking for a new game in town.
Mike: That fuels the demand. It’s the plain strategy of diversifying your ads, man.

00:04:10 - 00:04:43

Mike: Fair enough, and I think it’s a good strategy. The second thing is there are still hopes out there—maybe rightfully so, we will see—that it is driving incremental revenue, maybe even at better efficiency levels. Hopes are high because OpenAI is quite confident about the product.
Chris: For better or worse, rightfully so or not, we will talk about that in a minute.

00:04:43 - 00:05:02

Chris: But this is also driving demand—finally, ChatGPT is here, I can create incremental revenue, and maybe it’s performing better than Meta and Google.
Mike: I have my take on that. To regular listeners, it won’t be too surprising.

00:05:02 - 00:05:34

Mike: Let’s dig in a little bit further here. One thing I want to mention right off the bat is that this will include product carousel ads, also known pretty much as shopping. There are product carousels, which are a bit different on Facebook and Instagram, but we’re dealing with product listings in various formats. That means you’re going to need to set up the equivalent of a Merchant Center feed with OpenAI.

00:05:34 - 00:06:05

Mike: They have every interest in making that as easy as possible for people who are used to using Google, so I don’t expect any major hurdles there. It should be easy to set up.
Chris: But the ad platform is currently not self-serve. That’s the other side of this, right? What does that mean? Right now, there are limited partners, agency partners, and ad tech partners that they are working through.

00:06:05 - 00:06:38

Mike: I don’t truthfully know why or if that’s even necessary. I get that that was necessary earlier in the US when they were first building the user interface. To me, it feels like kind of a kickback thing—maybe they have an agreement or a contractual thing with some of these partners.
Chris: I don’t know how openly we can talk about that, but to a certain degree there’s an incentive for these certified partners to drive it.

00:06:38 - 00:07:00

Mike: Yeah, revenue share.
Chris: They’ll probably have partnership programs in place. They would be stupid not to do it that way, because all big platforms have partnership programs in place. But right now they’re narrowing the market at first by not making it self-serve.

00:07:00 - 00:07:18

Mike: That’s true. By narrowing it, they can make sure these limited partners are driving share of wallet. But like you said, maybe the main reason is that it’s a tough space to be in with all the regulations. They want to move deliberately and carefully to control it in a better way.

00:07:18 - 00:07:37

Chris: We’ll find out how deliberate and careful, because they’ve been moving fast. What did Meta always say? “Move fast and break things.” I think they’re going to keep moving fast.
Mike: They announced this August 18th and said self-serve will be coming later this summer.

00:07:37 - 00:07:54

Mike: Later this summer! At the time of recording, it is later this summer. Don’t be too harsh on Sam, there’s still some time.
Chris: Define summer. I don’t want to get into semantics, but we’re running out of time here. Should I write him on LinkedIn?

00:07:54 - 00:08:16

Chris: To all the listeners out there, that means if I’m an online retailer with a strong performance marketing team and I want to run ChatGPT ads on my own, I can’t do it right now because it’s not self-serve. I need a certified partner.
Mike: Yeah, but this will change probably very soon.

00:08:16 - 00:08:40

Mike: The exception here, from my understanding, is that people who have access to the self-serve platform in US markets can already target here. International advertisers are able to target in a self-serve way.
Chris: That’s interesting.

00:08:40 - 00:09:03

Chris: When that UI comes… we’ve talked before about some of their features and capabilities, but let’s take stock of where they are.
Mike: A feature-by-feature comparison, I like that! Can we pull up that slide?

00:09:03 - 00:09:23

Mike: Let me put this on screen for viewers. We have a table, a little cheat sheet here so we don’t lose track of anything. Let’s go through it. This is neutral—just a description of a feature-based comparison without bias. This is what it is.
Chris: Exactly.

00:09:23 - 00:09:40

Chris: What I also want to put on Front Street is that OpenAI is moving fast. The question is, are they moving faster than Meta, Google, and co. because they actually want to gain ground on them?
Mike: That’s the big question.

00:09:40 - 00:10:09

Mike: At the end, we can wrap this up into how this boils down into expectations. Circling back to that high level of demand, happiness equals reality minus expectations. If your expectations are high, that’s the thing I worry about a little bit.
Chris: Do you want to kick us off at the top here, Chris?

00:10:09 - 00:10:31

Chris: Yes, of course. To all the listeners, we have a table summarizing the most important features. Let me start with keyword targeting in ChatGPT ads: it basically doesn’t exist. In Google Ads, you still have your targeting possibilities, although Google is moving towards broad match, but you still have a control layer. That’s a massive advantage on Google’s side.

00:10:31 - 00:11:02

Mike: ChatGPT has contextual hints because a lot of the advertising is contextual. They give a hint about which context to look out for. Google has a similar feature in Performance Max with search themes. Very comparable to search themes, so there’s nothing new there. Google has search themes and various flavors of keywords.

00:11:02 - 00:11:30

Chris: Expectation management is all about managing expectations. When you start with ChatGPT, one thing is for sure: you have way fewer features, functionalities, and controls. That’s pretty clear, especially on the keyword targeting level.

00:11:30 - 00:11:55

Mike: Real quick thought here: I wonder if they will ever add keyword targeting. Google is moving away from it, and ChatGPT maybe fears that if they add keyword targeting, it’s going to read just like Google Search.
Chris: Yes, and they want to deliberately be a different product.

00:11:55 - 00:12:19

Mike: You’re right, maybe there will never be this control layer. It is what it is. We’ll find out.
Chris: Next up: demographics, for example age, gender, income. This is currently not available in ChatGPT ads, whereas in Google Ads, even in PMax, you have an increasing level of control over this.

00:12:19 - 00:12:44

Chris: For advanced online retailers—multi-brand retailers, B2C players who know who they want to sell to—this feature is perceived as important. I would recommend ChatGPT have a look at this because people are certainly interested in it.
Mike: Definitely, so important for building audiences.

00:12:44 - 00:13:06

Mike: Geography—should we go next there?
Chris: Yes, of course.
Mike: In ChatGPT, you can select at a country level. You can choose, for example, this beautiful green country with mountains that we live in, Austria. That’s the level of targeting.

00:13:06 - 00:13:31

Chris: You have to start somewhere, right?
Mike: In the US, they already have state level and postcode level, so they are heading in that direction. But in comparison, Google has all manner of geographic targeting options, all the way down to radius-based targeting.

00:13:31 - 00:13:55

Chris: For most online players, targeting at the country level is enough. But there are so many use cases where you actually want to be way more granular, so they still have to play catch up.
Mike: Especially for the SMB market where businesses are local. For national advertisers, country level is fine.

00:13:55 - 00:14:11

Chris: If we do a quick pause here, it shows again that even with their bandwidth, smartest people, and truckloads of money, it’s just not that easy to build an ad platform.
Mike: It’s one thing to have a great LLM, but building an ad platform is not easy.

00:14:11 - 00:14:32

Mike: They’re moving at breakneck speed, but it still takes time to reach parity.
Chris: The expectation in terms of revenue driving is one thing, but ChatGPT has to be this new kid on the block. What is different? Where is the advantage?

00:14:32 - 00:14:56

Mike: Thinking about it, Chris, the elephant in the room right now is that ads are not live in AI mode in Europe right now for Google. ChatGPT ads are inherently interesting because you cannot appear in Google’s equivalent yet.
Chris: So it is the only opportunity to appear next to AI overviews in Europe right now.

00:14:56 - 00:15:21

Mike: That is the opportunity right now in Europe. I promise you, Google is going to close that gap fast, probably in Q3 or Q4, because ads in AI mode are low-hanging fruit for Google.
Chris: They have done a lot of preparation work there and are scaling in the US. It will come for sure.

00:15:21 - 00:15:50

Mike: Next up: Customer Match. They work with hashed email and phone numbers, quite standard. But you need a pretty big list—about 25,000 matches—which might be a dealbreaker for a lot of advertisers.
Chris: That’s a lot! For the SMB segment, that takes time.

00:15:50 - 00:16:18

Mike: That has also been Google’s challenge over the past years: penetrating the SMB segment where a lot of growth occurs. A 25,000-match minimum might be a neckbreaker here. With Google, there’s no minimum.
Chris: Remarketing and retargeting: not available as of now.

00:16:18 - 00:16:43

Mike: They’re going to find a way to solve that, they have to. In Google Ads, you have full capabilities there.
Chris: Same for lookalikes and similar audiences. Lookalikes are hugely popular on Meta and available via Demand Gen as audience signals in Google.

00:16:43 - 00:17:11

Mike: I would love if ChatGPT could stand out by bringing these on as true targets rather than just signals. Google users complain that you cannot use remarketing lists purely in a targeting fashion.

00:17:11 - 00:17:33

Chris: 100%. We should think about where the differentiated value of ChatGPT could be. Down the line, even if features are missing, performance is what matters. If performance is there, clients won’t care as much about missing features.

00:17:33 - 00:17:53

Mike: Smart Shopping was a shitty product initially, but it worked fundamentally, delivered what it promised, and got adopted. Performance Max did the same.
Chris: Exactly. Bidding models—that’s a big one.

00:17:53 - 00:18:29

Mike: They have CPC and are working on conversion-oriented models. With Google, the big differentiator is Target ROAS (tROAS).
Chris: I don’t think Sam Altman ever imagined he’d one day be running a tROAS platform, but Sam, you’re going to be running a tROAS platform! It’s only a matter of time.

00:18:29 - 00:19:00

Mike: You have to! If we played the desert island game, my GOAT KPI is ROAS. Bidding models are directly connected to underlying performance.
Chris: Everyone will give ChatGPT a shot because of FOMO and high demand. But if ChatGPT is not producing after three months, it’s hard to bring advertisers back.

00:19:00 - 00:19:21

Mike: Once people turn their back, there’s huge pressure. The big challenge for them is: do we actually have good ROAS on this channel? If they don’t, they might put it off because if people see poor ROAS, they’ll reject it.

00:19:21 - 00:19:43

Chris: It’s a tough space to be in as a third or fourth player coming in. There’s not a lot of waiting room or goodwill. If you’re not performing as a platform, you will have massive trouble gaining market share.

00:19:43 - 00:20:10

Mike: It’s purely FOMO-driven right now. You hear pretty negative experiences out of the US. They really have an uphill battle. They should enjoy the FOMO while it lasts because the next phase will be the honeymoon ending, where they have to fight for every penny of budget.

00:20:10 - 00:20:42

Chris: Let’s look at measurement. It’s pixel-based with a conversions API, similar to Meta’s model. But it’s got up to a two-day lag in it, which sucks.
Mike: Google operates with Google Tag, which is on almost every website. It’s near real-time, populates fast, and has enhanced conversions.

00:20:42 - 00:21:05

Chris: Google has invested so much in data interpretation models to prove the value of Google campaigns. Next to bidding, ChatGPT has to offer measurement that proves the revenue they bring is at least incremental.

00:21:05 - 00:21:26

Mike: We know click-through rates are bad on these ads right now. This might not turn out to be a high-ROAS channel; it might be more mid-funnel, and they’ll have to frame and educate around that.

00:21:26 - 00:21:46

Chris: Effectively, all ChatGPT users also use Google. An ad might get clicked on ChatGPT, but then another ad gets clicked on Google where the final click happens. Google gets the attribution!
Mike: Google is one of the biggest referral beneficiaries. About 1 in 4 ChatGPT sessions immediately leads to a Google search where the final click happens, which is money in the bank for Google.

00:21:46 - 00:22:11

Chris: Understanding the true value of the channel must become priority number one. Every platform will claim revenue for themselves.
Mike: That’s why I strongly believe in Marketing Mix Modeling (MMM). You need objective, neutral technology to tell you the true value of each channel.

00:22:11 - 00:22:34

Chris: Just yesterday, we ran an MMM test run for one of our biggest clients spending heavily on Meta. Meta was claiming very little revenue, but MMM proved Meta contributed significantly more to bottom and top line than Google.

00:22:34 - 00:22:51

Mike: That measurement layer is so crucial. With ChatGPT joining the game, measurement will be a huge focus topic for them as well.

00:22:51 - 00:23:05

Chris: Product feeds: you have the ACP feed specs.
Mike: Their feed specs are good! Google actually copied some of what they did, so shout out to OpenAI. They did a fine job with feed specs, and I don’t foresee major problems there.

00:23:05 - 00:23:30

Chris: The advantage of Google is how many surfaces are powered by feeds—PMax, Demand Gen, YouTube. They are feed-based distribution beasts.
Mike: OpenAI needs to do the work, but data feed-wise, they are closest to feature parity with Google.

00:23:30 - 00:24:19

Chris: Last but not least: campaign management. It’s not self-serve yet in Europe, and there’s no public API.
Mike: For many advertisers, an API is crucial. I can’t wait for an API—I’m going to have fun with that thing!

00:24:19 - 00:24:39

Chris: For the SMB market, an API isn’t relevant, but ChatGPT wants the big advertisers who look at APIs carefully. What are the key takeaways, Mike?

00:24:39 - 00:25:03

Mike: Advertisers are frantic for this opportunity because of FOMO. There’s a valid reason: a huge user base. These ads will be served to free users. Media properties like Netflix or YouTube serve ads to free users, so that’s fine.
Chris: But there’s no evidence that CTRs or CPCs will offer cheap arbitrage.

00:25:03 - 00:25:21

Mike: Demand is high, and the platform isn’t fully optimized yet. Approach it in the spirit of testing. If you expect it to work right away, you’ll be disappointed. Reality minus expectations!

00:25:21 - 00:25:40

Chris: People arrive at the party early, but don’t leave before the party really gets started. Three months isn’t enough time to judge ChatGPT ads properly. Stay at the party, ensure you have proper measurement technology to evaluate contribution, and manage your expectations.

00:25:40 - 00:26:00

Mike: Agreed. We were fair to ChatGPT, and Sam Altman… I’ll bite my tongue! Thanks everyone for listening. This has been another episode of Growing Ecommerce, brought to you by Smarter Ecommerce (smec). Visit us at smarter-ecommerce.com. We appreciate your YouTube comments, LinkedIn posts, and reviews!

00:26:00 - 00:26:16

Chris: See you next time!
Mike: On to the next one.