Ads already follow you wherever you go. They’re on your TV, your phone, your train car–even on your airline tray table and escalator.
Now, they’ll soon be in your chatbot, too. OpenAI announced last week that it will begin selling ads in ChatGPT.
The move opens up a potentially massive revenue source for OpenAI–and is a huge threat to Google’s world-dominating ad empire.
Here’s why.
ChatGPT, Sell Me a Toaster
For years, OpenAI has resisted the siren song of advertising, and has kept its chatbot largely open to the world. That’s gone well for the company.
Offering a massively valuable product for free has been, unsurprisingly, popular. ChatGPT now has a reported 900 million monthly users.
Of those, 95% are on the company’s Free or Go tiers, which means they either pay $8 per month for the service, or pay nothing at all.
Providing cutting edge AI to a tenth of humanity, though, is exquisitely expensive–OpenAI expects to burn through $115 billion in the next few years.
To raise that kind of money, OpenAI needs to prove that it can monetize its vast trove of free users. Advertising, traditionally, has been the way to make money from non-paying eyeballs.
Indeed, in its announcement, OpenAI confirmed that its ads will initially be limited to Free and Go users. They’ll appear below organic answers, and will be specifically identified as advertising.
The company says that the content of ads won’t determine the answers that ChatGPT gives–if you ask it for a toaster recommendation, for example, it won’t write glowingly about a toaster brand that happens to be an ad partner.
But if you ask the bot how to fix your broken toaster–and it turns out your current unit is a lost cause–ChatGPT might present you with paid ads for a replacement.
A Better Mailbox
Showing contextual ads beside organic search results is hardly new–that’s been Google’s business model for decades, and it makes the tech giant hundreds of billions per year.
What’s different, though, is how OpenAI can target its ads.
Matching ads to a specific user is hard. Google has traditionally done it by gathering vast troves of data about all of us, and then mining that data for insights on what we might buy.
In some cases, it’s obvious what ad Google should show. If you search “flights to Aruba,” you’re probably traveling to Aruba, and the company can profitably show you a bazillion ads for Aruba hotels, Aruba rental cars, and the like for the next several months.
But more than half of Google queries are “navigational.” These are purely transactional phrases that people type into the search engine in order to find something else. Many more queries are short and ambiguous.
Looking at my own recent Google search history, for example, I typed in queries including “Sonicare skip slow start”, “Rancho San Ramon Community Park” and “what minor American celebrities are big in Japan.”
Good luck figuring out what ads to serve me based on that.
In contrast, when people chat with ChatGPT, they tend to do so for a long time. The average “session” with the chatbot reportedly lasts 12 minutes and 24 seconds.
That’s long enough to exchange a lot of information. And all that information gives the bot a strong sense of what the user wants–or might want to buy.
In a recent session with ChatGPT, for example, I had a detailed conversation about whether a specific piece of lumber was pressure treated, and how this might affect my ability to use it in my laser cutter.
From our long chat, the bot now knows the exact project I’m working on (replacing a mailbox). If ads were live, it could use that context to sell me all manner of related things–better lumber, metal mailbox bases, construction adhesive, and the like.
And it gets even better. ChatGPT knows I’m building a mailbox. But with its Memory function, it remembers from our previous chats the exact model of laser cutter I have, where I live, and much else about me.
That would potentially allow the bot to serve me ads not just for lumber, but for lumber that works in my laser cutter, would hold up well to the climate in California where I live, and would be easy for me to work with, given my exceedingly limited patience and skill for woodworking.
A Shot Across the Bow
OpenAI’s exquisite knowledge of its users’ needs means it can likely sell ChatGPT ads for a premium. And because so many people use the bot, its potential ad market is massive.
That makes OpenAI a huge threat to Google, the dominant player in the digital advertising market.
The search engine processes about 8.5 billion queries per day. Since the average person searches Google about 4 times per day, that means Google Search likely has around 2.5 billion active users.
That’s roughly triple OpenAI’s user base. But OpenAI is still growing aggressively. And again, the company’s users spend far longer with ChatGPT than Google’s users spend on its search engine–12 minutes versus as little as 5 seconds for a Google search.
The combination of a gigantic audience and lots of engagement time–plus a huge amount of targeting data–means OpenAI has all the tools it needs to threaten Google’s ad dominance in very short order.
And there’s precedent for a new entrant to the ad world gaining ground quickly. In 2019, Amazon had only $12.63 billion in ad revenue. But 2024, that number swelled to $56.21 billion and is still growing. And that’s with far fewer users than OpenAI.
Make no mistake, then–OpenAI’s announcement is a shot across Google’s bow.
The company says its ad program is still in a testing phase. But so was ChatGPT when the company launched it.
That OpenAI product completely changed the fabric of the digital world, altering how AI is built and sold. OpenAI’s ads could soon do the same for the way it’s monetized.