OpenAI wants to be Google. The company would likely reject the idea publicly, but think about the economics of advertising. Google built one of the best businesses in history by putting companies in front of people at the moment they say they need something. People forget how big that business is.
Google generated nearly $295 billion in advertising revenue last year. Search and other properties alone produced $225 billion—about $4.3 billion a week. The Trade Desk, the largest independent demand-side platform, booked $2.9 billion in annual revenue. Google’s search business generates more revenue in an average week than The Trade Desk does in a year. ChatGPT now sits at the same moment of expressed intent that Google has monetized so effectively. Whatever OpenAI calls its ad business, that is the prize.
CHATGPT SHOWS INTENT
Consider how you use tech products. You open Instagram to lean back and scroll. You don’t need anything, but maybe something catches your eye. You open Google because you need a new blue sweater. That’s high intent.
Specific searching is the way in which many people use ChatGPT. This is the important part. Nobody types “sweater” into ChatGPT. They say they need something warm for a week in Chicago in January. They add that wool makes them itch and that they’d rather not spend more than $150. Those prompts become a brief. We hand over our wishes, must-haves, constraints, and worries. Then we compare options and decide.
That comparison is where every sweater brand wants to be standing, and OpenAI has started selling the spot. Advertisers can buy on a CPM or CPC basis through OpenAI’s beta self-service Ads Manager, measure conversions with a pixel and Conversions API, and access campaign-level performance reporting. OpenAI is also testing Sponsored Agents, which let a user start a separate conversation with an advertiser after clicking an ad.
So the plumbing exists.
A SEARCH-CALIBER AD PLATFORM
But what made Google an unrivalled profit machine is that the advertiser can see the demand. I pick the keywords I want. I get a report of the searches that triggered my ads. I see which searches turned into customers and which burned money, and I block the losers with negative keywords. ChatGPT gives me nothing comparable yet. My ad shows up next to a conversation, and I have no idea what kind of conversation it was.
OpenAI needs to build analytics that show advertisers the consumer intent based on their prompts. I want to know which categories of intent my impressions ran against. I want to know whether those people were exploring, comparing, or ready to act, and which needs and product criteria drove that engagement. I want to know which kinds of sweater prompts turned into business and, just as important, which didn’t. Google’s advertisers have had some version of that map for two decades. It’s why brands keep paying.
Of course, privacy is always a concern. People might be telling ChatGPT things they would never type into a search box, and no advertiser should get to read a conversation or see an individual prompt. Fine. Nobody is asking for that, and it isn’t an excuse. Google has worked around the same problem for decades. It limits, aggregates, and buckets the queries too rare to show, and advertisers still get what they need. The solutions are well known. OpenAI just doesn’t have the system—yet.
LOOK FOR WHAT LEADS TO RESULTS
The ad industry already turns web behavior into pseudonymized, aggregated signals about audiences and intent. OpenAI can build a stricter version of that for conversations. It already says advertisers get aggregated performance insights with no access to individual chats. Apply the same rule to intent: Roll up the commercial patterns inside conversations into categories and keep the conversations locked. We don’t need to know who said what. We need to know which kinds of needs, at which decision stage, are producing results.
This matters because of how budgets get assigned. A CMO funds a platform based on what it can prove. If ChatGPT gives me reach, impressions, and clicks with no real view of intent, I’m going to file it under experimental or awareness. The money then comes out of the innovation line or paid social. If it can connect privacy-safe intent categories to business outcomes, it gets to compete for paid search dollars, the budget set aside for capturing existing demand. That is the money pile.
MAKE INTENT MEASURABLE
Marketers have a say in which pile OpenAI gets to chase. If you want ChatGPT to become a search-caliber channel, buy it like one. Run it against paid search as well as paid social and see which it behaves like. Ask your rep what intent reporting exists today and keep asking. A click on a ChatGPT ad tells you somebody was curious. Until OpenAI shows you the intent behind it, don’t pay search prices for it.
OpenAI already has the hard part down, the thing Google spent years teaching people to do. People are telling ChatGPT their goals, problems, preferences, objections, and purchase criteria in their own words. What it hasn’t built is the advertiser’s map of all that intent.
Building it is good for the rest of us, too. When advertisers understand what people want, the ads get better. Someone who asked about an itch-free sweater under $150 should see one. That’s an ad most of us would be glad to get from a chatbot.
If OpenAI wants to be Google, it can’t hide the data for much longer. It has to make that intent measurable without making it personal.
Patrick Briggs is CEO of Semify.