With each new detail OpenAI reveals about ChatGPT’s nascent ads business, more questions and potential challenges arise.
The first challenge stems from how new ChatGPT ads are. Advertisers may be reluctant to put dollars toward an untested channel, especially one that – unlike its competitors – doesn’t offer a pixel or any sort of attribution system.
At first, it seemed like OpenAI was planning to head off potential advertiser hesitation by simply charging a lower cost; early advertisers were asked to put less than $1 million toward ad spend during the initial launch.
But hardly a week later, OpenAI announced plans to charge a $60 CPM, comparable to the cost of running ads during an NFL game – and more than three times the CPMs on Meta.
Now, however, the bigger question isn’t about pricing or what the ads will look like, or even when ChatGPT will become more widely available to advertisers.
It’s how OpenAI will convince buyers to test this new channel – and justify its reportedly sky-high cost – in an increasingly competitive advertising landscape.
The AI chicken or the AI egg?
One area where chatbots have a leg up on other platforms is contextual relevance.
At any point in a conversation, ChatGPT can target a user with an ad at the right moment based on the content of their current chat, said Yang Han, CTO of DSP StackAdapt. Other platforms might have an idea of what types of products to show a user (perhaps energy drinks and activewear for a runner), but it’s a lot harder for them to know exactly when to reach that person.
Crucially, ChatGPT doesn’t just have user data; it has memory. Beyond knowing that someone is a runner, it may also know exactly when they last ran a race as well as related queries they’ve made in the past.
ChatGPT could suggest not just a relevant product, but the version of that product that’s best suited for an individual consumer, Han said.
Becoming “fully contextually relevant” will take time, he said, since OpenAI is only working with a small pool of advertisers initially. Until ChatGPT has the same breadth of advertisers that, say, Meta has – and Meta works with millions – it risks “alienating users” by showing mediocre ads, which means new advertisers might be resistant to sign on.
It’s “a bit of a chicken-and-egg problem,” said Han.
Discovering one’s identity
But some advertisers are more comfortable than others with diving in headfirst.
Tyler Holoubek, assistant director of investment at independent agency KSM Media, has seen this firsthand. Some of KSM’s clients, he said, have been extremely “gung-ho” to test ads in ChatGPT and glean as many early learnings as possible. But others, especially those in more regulated industries like pharma and financial services, are far more hesitant. They have to ensure that specific compliance requirements, like the type of content they appear next to, are being met.
Brands have spent a lot of time perfecting their search ad strategies, but AI search is an entirely new beast – according to some, at least.
“ChatGPT is not Google,” said Nic Baird, co-founder of Koah, an ad network built specifically for generative AI platforms. Ten percent to 20% of Google searches are commercial queries, he said, compared to only 2% to 3% of ChatGPT queries.
In Baird’s view, the ads will need to get a little more creative regarding what sorts of queries they’re appearing next to.
For instance, Baird said, if someone is using ChatGPT to vibe-code a financial services app, they might receive an ad for a fin tech company to help speed up the coding process or provide integrations to the app. Giving users concrete value and solving problems on the spot is how platforms like ChatGPT can establish trust through ads, Baird said.
Search showdown?
However, others believe ChatGPT can become competitive in the ad space by taking the exact opposite perspective to Baird and positioning itself as the next phase of search.
OpenAI should go the “aggressive route,” said Holoubek, and frame its chatbot as a “direct competitor” to Google.
ChatGPT isn’t likely to replace Google Search, he said, but OpenAI will be competing for advertiser search budgets.
And although most ChatGPT queries aren’t directly commercial, Diaz Nesamoney, CEO of agentic marketing platform DaVinci Commerce (formerly Jivox), believes chatbots have a unique advantage in ecommerce.
ChatGPT could be a sweet spot for CPG brands looking to avoid middlemen and reach consumers directly, as opposed to going through a retailer, Nesamoney said.
An LLM model creates what he called a true “three-party transaction” between the consumer, the model and the brand, whereas retailers have often pushed for a “two-party transaction,” in which “the brand just happens to be there.”
OpenAI’s ad offering could be an opportunity to drive more direct sales via brand websites, he noted. Still, it’s unclear whether ChatGPT’s in-app shopping integrations will ultimately counteract this.
After all, ChatGPT’s goal is “to keep people inside ChatGPT,” said Phillip Thune, CEO of AI search intelligence platform Adthena. Which is why, he noted, charging per view, rather than per click, makes sense for OpenAI, because clicks that lead to an integrated shopping experience benefit the platform – but clicks that sends users away certainly don’t, he said.
In-platform purchases, of course, can be detrimental to brands, who benefit financially from direct sales, not to mention the data they get from first-party purchases.
But even if ChatGPT ends up limiting a brand’s ability to maintain a direct relationship with customers, it still has unique value over traditional search.
Holoubek is optimistic that the “semantic specificity” chatbots gain from engaging with users in great contextual detail will set them apart from other ad platforms, along with the option to engage directly with a brand agent, which OpenAI teased in its initial announcement.
The next hurdle, he said, will be understanding the “sensitivity” of a specific conversation and whether it’s “appropriate” to show an ad within a given context. For instance, if you’re asking for advice on how to cope with the loss of a loved one, you might not be jazzed to receive an offer for a mood-boosting exercise plan.
Determining the nuance of when ads are appropriate leaves a lot to consider. After all, robots aren’t exactly known for their sensitivity.
