Like a kid playing dress-up in their parent’s clothes, OpenAI is trying to look all grown up. But it still has quite a ways to go.
Earlier this week, OpenAI announced that its ads business hit a $1 billion annualized revenue run rate.
For those less trained in finance-bro vocabulary, “annualized revenue run rate” is basically a way of saying, “If we keep earning revenue at the exact same rate we are now, this is how much we’ll make in a year.”
The metric isn’t particularly helpful for companies whose revenue numbers fluctuate throughout the year, such as brands selling seasonal products. In fact, annualized run rates can be pretty misleading in that case, like if an ice cream brand made projections based solely on its July sales.
But for nascent businesses that are actively and consistently growing – like OpenAI’s ads biz – an annualized revenue run rate can be a useful projection of the coming year.
Still, while $1 billion is nothing to scoff at for a business that’s scarcely six months old, it’s still shy of the company’s earlier projections, which estimated $2.5 billion in ad revenue this year and $100 billion by 2030. (Emarketer, on the other hand, estimates that it will be at roughly $5.41 billion in 2030.)
So, AdExchanger asked advertisers what it would take to get them to spend more. And, as it turns out, advertisers want to see the same improvements they demand from every emerging media channel: namely, better measurement and accountability.
Getting your sea legs
While OpenAI’s run rate projection doesn’t match its earlier estimates, the company is clearly pulling out the stops to bring advertiser demand to its platform.
Many advertisers have received advertising credits from OpenAI over the past several months, meaning that they didn’t have to put actual ad dollars toward their ChatGPT advertising – it’s more like paying with a gift card.
But OpenAI clarified to AdExchanger that these ad credits are not included in its $1 billion run rate projection. So the figure is only based on money actually spent by advertisers.
Looking ahead, what will it take for $1 billion to turn into $2.5 billion, let alone $100 billion?
The short answer: better measurement capabilities and standardization.
Right now, a lot of clients see “value in being early” to ChatGPT ads, according to Monica Shukla, VP of biddable COE at agency Mile Marker. They can better position themselves and develop relationships with the channel “before the space gets commoditized.”
However, as the channel matures and more of OpenAI’s competitors launch ad businesses as well, “advertisers will start to more deeply weigh the ROI of the activity against its perceived value,” which will require “tangible metrics,” said Victor Batista, senior director of paid media at digital marketing agency Jellyfish.
One of the “tangible metrics” that is most obviously lacking at the moment is competitor data – something that brands are increasingly seeking out for benchmarking, according to Batista.
Currently, the lack of competitor data, as well as the “very limited conversion-tracking capabilities and audience options” makes the platform feel “very rudimentary,” said Rajeev Nair, co-founder and chief product officer of measurement platform Lifesight, which is an early ChatGPT advertiser.
OpenAI needs to be more compatible with, or actively support, API integrations with third-party MMM and incrementality tools, he added, if it wants budget owners to spend more confidently.
The platform’s introduction of a conversion API and the recent launch of conversion-optimized campaigns (which are geared toward a specific metric, be it reach, clicks or conversions) have been steps in the right direction, said Emil Dewy, senior manager of programmatic media at M+C Saatchi Performance.
But, Dewy added, because the generative AI search channel is seen as a mid-to-bottom funnel play, it needs to introduce new optimization and measurement tools “that support different performance goals.” Until the channel gives advertisers more flexibility for optimizing to different KPIs, he said, advertisers will still see it as “experimental/optional.”
Plus, Dewy added, ChatGPT is lacking in tools that measure prompt-level trends and broader behavior. And some other use cases that advertisers “take for granted,” like being able to suppress existing users from ads and instead only targeting new prospects, are currently missing from the AI search channel, he said.
OpenAI declined to comment on its road map for introducing new measurement solutions.
It just takes some time
Part of ChatGPT Ads’ growth is going to simply come from trust built in the channel over time – specifically in its ability to influence downstream traffic, rather than just immediate conversions, said Lifesight’s Nair.
One of the biggest draws of ChatGPT Ads for advertisers has been the opportunity to get in front of consumers at the exact moment they’re ready to purchase.
Conversational queries are a great way to “surface need-states and purchase intent,” and have been driving incremental sales for brands, according to Mile Marker’s Shukla.
Users often demonstrate stronger intent on ChatGPT than on standard display channels, Dewy agreed, and M+C Saatchi Performance has appreciated the ability to “skew spend towards conversations where [it sees] stronger commercial intent.”
Still, a lot of ads served to ChatGPT users aren’t responding to queries that are directly related to purchase intent, but rather more general queries. For example, a user asking for advice on how to structure a resume might see a Canva ad, or someone asking for recipe ideas might get an ad for a grocery delivery service.
The problem is figuring out when those ads lead to a purchase down the line, rather than a spontaneous, immediate conversion.
So far, standard ChatGPT queries – with no ad involved – have been proven to drive direct and organic traffic weeks or even months after a person has researched a brand on ChatGPT, per data analysis conducted by Lifesight.
There’s a good chance that ads in ChatGPT will have a similar effect. But until Lifesight has a bit more data under its belt (it needs about six months to accurately model cross-channel lift), “dollars will remain in ‘test’ budgets rather than moving into core allocations,” said Nair.
So, for now, it’s hard to understand the long-term value of the channel without a way to measure its long-term effects.
