← Behind the Numbers: an EMARKETER Podcast
Behind the Numbers: an EMARKETER Podcast
ChatGPT With a Side of Ads: Will AI Chatbots Ruin—or Reinvent—Search Advertising? | Behind the Numbers
00:27:02
Loading summary
Transcript53 lines
- [00:00]
A
Ever seen an ad that just fits perfectly? That's Seed tag. Their neuro contextual technology combines AI and neuroscience principles to place brands exactly where they belong. It's privacy first advertising that taps into the user's interests, emotions and intentions, making every interaction feel natural and relevant. Seed tag, where context becomes intelligence. Hey, gang. It's Friday, February 6th. Marissa, Nate, and listeners, welcome to behind the Limits new marketer video podcast made possible by ctag. I'm Marcus, and joining me for today's conversation, we have two New York residents. We start with our analyst, Marissa Jones. Hello.
- [00:42]
B
Hi. It's great to be back, Marcus.
- [00:44]
A
Yes, indeed. Welcome to the show in 2026. We're also joined by principal AI analyst Nate Elliot.
- [00:51]
C
Hello there, Marcus.
- [00:53]
A
Hey, fella. Thanks for being here. We start, of course, with the fact of the day.
- [01:01]
B
All right.
- [01:01]
A
Which film has the longest movie credits? So, according to the Guinness World Records, Avengers, Infinity War holds the record for the most credited individuals on a movie as listed on IMDb with how many people? How many people do you think are listed in the credits for that film?
- [01:19]
C
I mean, a masterpiece like that must have millions.
- [01:23]
A
It is millions. It's not millions, but it's high. Marissa, any guesses?
- [01:27]
B
Outside of me, you know, I'm not sure if like a guess like 300 is too high or way too low.
- [01:33]
A
300 people? Yeah, just 300. Have you sat through the credits of a.
- [01:37]
B
No, I hadn't.
- [01:38]
A
Okay, 300 minutes maybe lasts for a while. 5,109 people. Is that the cast and crew as January 2024. That makes the movie. That makes it the movie with the greatest number of credits currently record recorded. The largest contributor to this total is the film's enormous visual effects team, which consists of about 2, 700 people. The credits for Once Upon a Time in the west and also. What's the other one? Superman. That film also have the record for the longest credits ever. So they go on for the longest amount of time. 12 minutes. Those credits roll for. And as we've mentioned before on the podcast, the actual action of an NFL game when they're actually playing the game. Not that when they're in the huddle or when they're doing replays or when they're just down on the sideline is 11 minutes. So the credits of Superman are longer than the football game action.
- [02:45]
C
That's tragic. Bizarre comparison.
- [02:51]
A
This is gone. It's gone off the rails. That's where it's gone. Four people were involved in the making of this podcast, not 5004. Three of them on the show. The other is Lance in the background. Thank you, Lance. Anyway, Today's Real Topic ChatGPT with a side of ads. All right, talking about ChatGPT with ads. ChatGPT's free ride is ending. Here's what OpenAI plans for advertising on the chatbot, writes Matt O' Brien of the Associated Press. OpenAI will soon start testing ads by showing them to ChatGPT users who aren't paying for a premium version of the Chatbot. The ads will appear at the bottom of ChatGPT's answers when there's relevant sponsored product or service based on your current conversation, so they say. So searching ChatGPT for places to visit in Mexico could result in essentially a a holiday banner ad appearing. The company has made a commitment not to serve ads to what it believes are under 18 year olds and they say ads won't appear alongside sensitive topics like health or politics, most importantly Fiji. Simo, the company's CEO of applications, said most importantly, ads will not influence the answers ChatGPT gives you. Your ads will be influenced by your conversations. Personalization will be on by default with an opt out. Marissa, I'll start with you. What is your main takeaway from this chatgpt with Ads news?
- [04:11]
B
My main takeaway is that this is really where Jenny I is going to be able to keep up with the massive costs that go into making and operating a Gen AI platform. ChatGPT has been struggling for quite some time with profitability. It's very far from achieving that. And this is just to me, obviously the next step in becoming profitable. Except for people who can pay to afford the ads. Obviously that might increase subscriptions if they're people who don't want to see ads. So that's also a goal of it. But this is really OpenAI's best bet right now to inch toward profitability in the long term.
- [04:49]
A
Yes, same as me, of course, they added ads was what I thought first company valued at half a trillion, owes about a trillion on what they call future procurement commitments, contracts for AI chips and energy. OpenAI told investors in September they could burn over $100 billion in cash by 2030, according to the information in the Financial Times reported that in 2025 OpenAI operated at a loss at around 8 billion in the first six months of the year and that only 5% of ChatGPT's over 800 million users are paid subscribers. So they need money and quite quickly, Nate, it appears.
- [05:26]
C
Oh yeah, I mean of course they are offering ads. As you say. This was inevitable. This was always what was going to happen. And all the AI platforms have been protest years that they don't want ads, except for Microsoft, which has actually been running AI ads for like three years now. All the others have been saying for years, we don't want ads, we don't like ads. Sam Altman himself has said that he hates ads, calls ads in AI uniquely unsettling, and once referred to ads in ChatGPT as a, quote, last resort. And yet it was always inevitable that they were going to run ads because as Marissa says, as you say, they're burning money like there's no tomorrow. I actually have a higher number. I believe their spending commitments are up to $1.4 trillion. And there's no end in sight to the red ink over there. So this was always the only option for them.
- [06:22]
A
Yeah, and others have, you know, Google already introducing ads and AI overviews and AI mode. So it's not like they're the first. My next question, Nate, I thought, will this scare users away?
- [06:33]
C
I don't think it will. Not in any meaningful number. I mean, listen, some people are going to leave because of this, but I don't think that number is going to be significant. It is their greatest fear. I mean, if you look at the state of AI right now, while everyone assumes that everyone uses AI, the reality is the majority of Americans don't use AI in any given month. As you can see from the chart on the screen, which happens to be in my new report on ChatGPT ads available at all good local eMarketer.com the majority of those who do do use AI in any given month are more casual users for whom the switching cost is close to zero. And we see the AI user base continuing to grow. So if you add up all the new users coming into generative AI over the next five years, and you combine that with all the casual users who could switch from ChatGPT to Gemini or Claude or Perplexity tomorrow without missing a beat. That's a huge number of people. And that's what OpenAI is afraid of losing. They don't want their casual users to leave. They don't want new users into this space to decide to go elsewhere. The reality is, though, ads are already elsewhere as well. So as I mentioned before, Copilot has been quietly running ads for three years now. Google introduced ads into AI Overview about a year and a half ago. And just before ChatGPT announced their ads in January, Google announced their first AI specific ad unit, which are effectively sponsored coupons in AI mode. And I fully suspect that we'll see ads from Perplexity and Grok and Meta AI within the next year as well.
- [08:12]
A
Marissa, I mean Jeremy Goldman, who's one of our colleagues, was speaking to Reuters and he said if ads feel clumsy or opportunistic, users can easily switch. Not that they will, but can could easily switch to rival chatbots. What do you think of how this is going to sit with users?
- [08:31]
B
I do think there will be some resistance, but not necessarily enough to convince users to flee the platform en masse. Like Nate said, going to another platform if they don't already have ads, they'll probably be getting ads at some point. And more people are using ChatGPT than other platforms. It's kind of a lot of people's first thought when you think of AN F gen AI platform is ChatGPT. So I don't think this will be enough. And we also have to consider that users are very good at tuning out ads entirely. This for a lot of users might just be something they notice once and then they tune out for the rest of their experience because they do that with other platforms.
- [09:17]
A
How much more effective, if at all? I'm assuming they are. But how much more effective are chatbot ads compared to traditional search ones?
- [09:26]
B
I do think they can be very effective. There's some data from Microsoft which obviously Copilot is kind of the one that's been doing this for quite some time time Microsoft Internal data showed 153% lift in click through rates and a 54% improvement in user experience actually from running Copilot ads compared to traditional search. So there is great potential for this. It really does come down to how ChatGPT integrates these ads which we're still kind of yet to see. The format is still very nascent. Obviously it might require a test and learn approach. Microsoft, like Nate said, has been doing this for years so they've had time to kind of work out the kinks and make a more established ad platform. ChatGPT is just now getting into it, so there is a chance at first it might not be the most effective thing. Advertisers might not see the best results at first, but in time this is obviously going to become a core part of the ChatGPT experience. So with that its effectiveness will likely grow.
- [10:31]
A
I was reading a piece for this and it was saying about how the company's hinted at taking this step further already and talking about how static recommendations could eventually be able to be followed up on asking questions about advertising products directly within the conversation, transforming ads into Extensions of the AI interaction. Zooming out a little bit. Nate, could you walk us through how this type of an ad does look different from what people are used to and how powerful that could potentially be?
- [11:07]
C
Yeah, I mean, there are limitless possibilities for what they could do with advertising in this space. And the reality is Microsoft in its three years of running ads and Copilot have tried out, I think, three or four different ad formats. We've seen other experiments with other ad formats from other players along the way and we, we don't know what the format will be for these ads. I mean, by all means, the easiest version of this is what Google's been doing in AI mode for the past year and a half, which is just taking regular search paid listings and dropping them at the bottom of or next to an AI powered response. That might not be the most effective form of advertising. It might not deliver the best results. But from a platform perspective, there's a lot of appeal there because billions of dollars per year are already spent on these regular search paid listings from nearly every advertiser on earth. And it's really easy for Google, for instance, to just choose some of those, run them in their AI products and call it a day. They would flip a switch and make a ton of money from that. Right away. Microsoft could have done the same thing and has. That's actually one of the formats they've tried alongside Copilot, but they've tried a richer experience that they call showroom ads. They've tried some comparison ads where they have tables of comparison content for different advertising brands. They've tried a bunch of different formats. I fully expect that we'll see OpenAI try a bunch of different formats as well as they work to figure out what's going to be most effective in terms of that triangulation of what makes them the most money from the most advertisers, what delivers the most value to the advertisers, and what's most effective or at least, least offensive to their consumers and their user base. We don't know what it's going to look like because they don't know what it's going to look like. But it's testing. Yeah, I mean, it could be back and forth conversations with a branded chatbot. I mean, when Google introduced its coupon ads in AI mode in January, one of the things they announced alongside that was branded chatbots. And so you could see those coming together Google @OpenAI anywhere else. The showroom ads that Microsoft has in Copilot are interesting in that they expand into almost full screen multimedia experiences that are designed to replicate the experience of going into a physical store. Of course, they don't quite do that, but they're certainly a lot bigger and shinier and flashier than a simple text listing. So there are a lot of different ways this could go. And OpenAI is going to need to take some time to figure this out, which is unfortunate for them because as we've all said already, they are burning money like there's no tomorrow and they need to ramp this up as quickly as humanly possible.
- [13:55]
B
So they're charging and they're charging a very high rate and are going to have to convince advertisers that whatever they do will be effective. So they're going to have to do a lot of testing and very quickly to figure out what works best, what drives the highest return on ad spend and what's most effective for advertisers.
- [14:13]
C
Marissa, you don't want to pay super bowl ad prices for largely untargeted, unmeasurable.
- [14:19]
B
Ad that we don't know if it works, right?
- [14:21]
C
I mean, that sounds reasonable, right?
- [14:23]
A
Do we think, I mean, in terms of their rate, we saw Netflix, when they first started doing ads come in with a very high rate and say it's a premium spot and we're not going to have as many of them. Do you think that they'll be able to have something that's above market down the road, or do you think that they're actually going to come down the same way Netflix had to come down something closer to market rate?
- [14:43]
C
I mean, they might, they might be able to get a good price for it, but the reality is the price isn't determined by the seller. The price is determined by the value that the buyers are getting from it. And so they can quote whatever rate card they want. And apparently the rate card they're quoting right now is $60 per thousand, which is very, very high. It's really fascinating that they're quoting a CP instead of an action based price like cost per click or cost per conversion. And it's really interesting they think they can get that amount of money. Now I've argued that AI brand experiences are closer to branding than they are to direct response. That when you look at where in the customer journey AI conversations sit, they tend to sit really high up in that funnel. Discovery, maybe even before the point of discovery, we see increasingly that these chatbots are responding even to explicitly non commercial queries like why does my back hurt with responses that mention products? And so, you know, if that's the case, and these are brand ads, then CPMs actually make sense. And high CPMs may be justifiable. But to make that make sense to justify, you need to help brands measure it. And from what we're hearing, the measurement capabilities that they're launching this trial with are exceptionally rudimentary. They're going to need to do better at helping brands and buyers understand the value that they're getting from these ads if they want to charge those prices.
- [16:22]
A
Yeah, you mentioned the does my backer and it's getting a lot of health related questions. And health related questions can be particularly personal. I wanted to throw this at you guys. Miranda Bogan, Director of the center for Democracy and Technologies AI Governance Lab, saying introducing personalized ads starts OpenAI down a risky path previously taken by social media companies. People are using chatbots for all sorts of reasons, including as companions and advisors. There's a lot at stake when that tool tries to exploit users trust to hawk advertisers goods. And Megan Marone and Kerry Flynn of Axios also shared some concern, writing that, quote, ads in chatbots were always inevitable and concerning given how much sensitive personal and emotional information people share with ChatGPT and other bots. Close quote. Nate, I'll start with you on this one. Do they have a point? Is there more of an ethical component to advertisers advertising during a search conversation versus traditional blue link format?
- [17:16]
C
I mean, sure, they have a theoretical point, but do they think that search engines and social platforms don't know anything about their users? I mean, sure, people type sensitive questions and content into AI models, but guess what, you've been googling stuff every bit is sensitive for probably decades now. Especially you, Marcus. I know you like to ask personal questions.
- [17:40]
A
We talk about that, please.
- [17:42]
C
And you've had very sensitive conversations and posted lots of personal information on social platforms as well. All of these sites, therefore all of these companies have incredible amounts of very sensitive information about us that we wouldn't really like to have exposed for public consumption and that we probably don't want use to target advertising. And the responsible companies in this industry are at least relatively safe in terms of saying we're not going to use things. If you declare a medical diagnosis or condition into Google, they're not going to use that to target ads. Back to you. Having said that, if you ask a question about medication in Google, by all means they're going to sell ads to competitors of that medication. Now I expect that we'll find something similar in ChatGPT ads and ads across other AI platforms. Having said that, again, we've noted that OpenAI is particularly desperate for revenue at this point. And OpenAI's executives have never shown an inclination to be terribly sensitive or responsible about matters like this. So we may see something different from ChatGPT, but this isn't new. This is, you know, this is just the latest technology, the latest set of technologies that have information about us that we wouldn't want to tattoo on our foreheads. And they follow in a long line of lots of other technologies and companies that also have similar information about us and also run successful, responsible advertising businesses.
- [19:17]
A
Yeah. Marissa, what's your take here?
- [19:20]
B
I would say I mostly agree and I think that's because while I do think AI can hold disproportionate weight for people who are. I've been doing a lot of research and writing on AI companions lately. So people who use these for companionship and really develop a personal relationship that they feel obviously in that case, if an AI is recommending you and you see this chatbot as your friend, even it might hold disproportionate weight. I do think Chachi Beatty, by saying they won't be serving anything harmful to minors, it could kind of curb some of those concerns. A lot of AI companion use happens among minors. A lot of teens have used AI as companions. So I do think that is an important safeguard that could help them avoid some of these problems that they would otherwise maybe be facing.
- [20:09]
A
Yeah.
- [20:10]
C
Having said that, this is the same company that claims Sora 2 wouldn't violate copyright.
- [20:14]
B
That's true.
- [20:15]
C
You can go through a long list of things that they've claimed they would do responsibly and well and then just launched whatever they had and that's what we got.
- [20:24]
A
Yeah, I like you wrote this line. Russia. I like it. It sums up this quite nicely. I think emotionally adept AI can strengthen engagement and foster brand loyalty on the good side of things, but missteps could lead to perceptions of manipulation. So, yeah, it depends how on the execution. Let's end by talking about the marketers here. So, Matt O' Brien of the AP writing, OpenAI gears up to launch ChatGPT ads. Marketers try to keep up. And John Mew, CEO at IAB UK, saying, quote, for marketers, the key issue is less about specific formats in the short term and more about whether conversational interfaces can support long term brand building alongside performance outcomes. Close quote. Marissa, I'll start with you. What should marketers do with this news that Chat GBT will have ads?
- [21:11]
B
I think the really only route is approach these with interest, but approach them with caution. Going too far into investment, especially with the prices ChatGPT is charging, especially with them not having any proven results yet. It's risky, but there is interest in this format. And ChatGPT is particularly well positioned because of its user base and because of how engaged a lot of its users are. But I really think the imperative is to just approach it as you would any new ad environment that is still in very early stages. Like I said, the user base could make it a really great platform for delivering ads to very engaged users. But we don't know fully the full picture of how privacy compliant these ads are, how transparent they will be. So it is still difficult to tell whether this will be a worthwhile investment. That helps brand building, that helps brand loyalty. But AI ads at the same time are really the future. I mean, AI is obviously the future and advertising is the next step in these platforms. So marketers do need to be paying attention to how this evolves and to know and pay attention to when this format proves its worth, which it eventually will, even if it's not with ChatGPT.
- [22:27]
A
Nate, in terms of advice and marketers, you went and wrote a whole piece on this in a ridiculously short amount of time. It's called the Gradual Then Sudden Rise of Ads in AI subscribers.
- [22:40]
C
What a bar title, huh?
- [22:42]
A
Get the full report? Of course, I will put the link in the show notes, but for folks who haven't read that yet, or folks who have and something else you want to say to them, what do you think marketers should be paying attention to here?
- [22:55]
C
Yeah, I think marketers should respond to the launch of ads in ChatGPT by buying ads in Copilot. I mean, there are a lot of challenges that OpenAI will have as they try to launch these ads. They don't have an established technology in place to manage the ads. They don't have established teams in place to help sell the ads or service the advertisers. They don't know their best format and haven't tried even a single format as far as we know at this point. They are charging incredible amounts of money, as we've discussed, and they're not providing very much in the way of reporting ads in OpenAI or on. OpenAI look like they're going to be difficult to turn a profit on, at least in the near term. But ads within Copilot actually don't have any of those problems. They have established teams and technologies, they've tried different formats, they have experience. The prices aren't ridiculous. They have reporting. If you're really interested, I would make sure that you're at least running ads in Copilot as well as on ChatGPT. But having said that, if you trust your paid search team, if you as a CMO or a marketer trust your paid search team, I would give them the keys to AI advertising and say spend as much as you want on these platforms. Because the thing that good search marketers do so well is they measure and they optimize and they update their investment levels in different keywords, different engines, different platforms and services from day to day, sometimes from hour to hour, to get the best return on investment. I said before, I don't think that ADS and ChatGPT are going to generate a great ROI, at least not in the first several months. Maybe the first couple of years. I mean, it's hard to remember this, but it's took Facebook several years before their ads stopped being terrible. Facebook ads were really, really bad when they launched. They had a lot to figure out. OpenAI is in the same position. They haven't done any of this before. They need to figure all of this out. There are other companies where they have done this before. And so I would say the AI specific ad placements that Microsoft is offering that Google's about to trial in AI mode, I would say those are equally valid places to trial ads in AI chatbots and more likely to be easier to work with and maybe produce better results. But as long as you're measuring and optimizing and only spending what you're getting your money back on, only spending as much as is delivering more value than you would be able to get from that dollar spent somewhere else. Then go for it. Spend whatever you want. I just think that if you follow those rules, you're going to end up spending not a ton in AI advertising in the near future.
- [25:45]
A
There are folks who are skeptical of this, one being Mark Ritson wrote a piece in Adweek titled OpenAI's ad offering as a last resort and it still won't save the company to compare it to Netflix again, we're talking about ad prices when they launched. Advertising didn't go well out the gate this year, they said, or 2025. They said they made one and a half billion dollars. That's the drop in the bucket compared to the 4050 that they're making. But they're eventually getting there. So this could absolutely take some time if it gets there at all. We shall see. That's what we've got time for. For this episode thank you so, so much to my guests. Thank you. First to Marissa.
- [26:22]
B
Thank you so much for having me.
- [26:24]
A
Yes, indeed. And of course to Nate.
- [26:26]
C
Cheers, mate.
- [26:27]
A
Yes. And thanks so much to the whole. The whole production crew. It's Lance. Thank you to Lance, John and Danny. Didn't do anything on this one. Lazy. And to everyone for listening to behind the Numbers, E Marketed video podcast made possible by ctag. Thank you to you guys. Subscribe Follow if you have some time, leave a rating review. If you have some extra time, we'll be back on Monday, happiest of weekends. Not to you, John or Danny. And do anything.