OpenAI is building a new kind of ad machine. Early data suggests Google should worry.

OpenAI's move into advertising is starting to look like the foundation of a massive new revenue engine, one that could challenge Google's search empire in significant ways.

That's according to new data from Similarweb executive Heral Amir, who's been studying early ChatGPT ads bought by hundreds of companies, including HubSpot, Dick's Sporting Goods, Nordstrom, Cursor, and Indeed.

He says the reason is simple: ChatGPT ads aren't built around keywords. They're built around conversations, and what Amir calls "conversational intent."

"When search was dominant, advertisers bought intent through keywords," Amir said. "Conversational AI changes that completely. The user's intent evolves during the conversation itself."

That shift may sound subtle, but it has enormous implications for how AI ads are targeted, priced, and monetized.

Ads appear later

Traditional Google Search advertising depends on users typing commercially valuable phrases like "best running shoes" or "cheap flights to Miami." ChatGPT, by contrast, can infer intent from a sprawling back-and-forth exchange, even when the user never explicitly searches for a product.

A user who never typed "headphones" can end up seeing a headphone ad, for instance. The context of a rich conversation creates an ad opportunity, not a Search query with specific keywords, according to Amir's analysis.

One user asked ChatGPT for help preparing for a job interview. Those conversations went back and forth 30 times before OpenAI showed an ad from Indeed.

Another user researched golf clubs began with broad questions before drifting toward specific product comparisons and purchase decisions. That involved 10 turns of the conversation with ChatGPT, before an ad from Dick's Sporting Goods appeared.

In another example, a ChatGPT user asked: "Where are distilled acid oils exported from the USA?" There were 52 turns of that conversation before an ad from HubSpot appeared.

"Intent drift"

According to Similarweb's data, 46% of ChatGPT users who eventually saw an ad started with no commercial intent. In other words, the conversation itself generated buying intent. Amir calls this "intent drift," the way a user's intent evolves as a chat continues.

That's a powerful advantage over Google's keyword model, which largely captures intent only after users explicitly express it.

"This is a real threat on Google's side," Amir told me.

Similarweb found that 83% of ad-triggering queries inside ChatGPT would never have activated a traditional Google Shopping ad. ChatGPT is effectively monetizing a layer of user behavior that search engines often miss.

"This format doesn't exist on Google or Bing, and it's a big opportunity," Amir said.

Valuable inventory

The structure of ChatGPT ads also differs sharply from Google Search. Instead of a crowded page full of ads from different brands all competing against each other, ChatGPT currently displays a single ad per conversational turn.

"No side rail. No competing cards. No ten blue links," Amir said in a recent presentation. "Just a single branded moment with a user who's already deep in the conversation."

That exclusivity is already making ChatGPT ad inventory valuable. Similarweb estimates ChatGPT ads may command about $60 per thousand impressions, known as CPM. That's a lot higher than Meta and LinkedIn ads. On a cost-per-click, or CPC, basis, marketers are paying about $12 for ChatGPT ads, well above Google Search, according to Similarweb estimates.

Engagement metrics

The early engagement metrics are turning heads, too. ChatGPT ads currently generate an average click-through rate of roughly 0.68%, according to Similarweb data, placing them between display ads and traditional Search ads.

The best performing ad campaign Amir spotted on ChatGPT had a CTR of 5.4%, above the Google Search average and in the top tier for social-media ads.

ChatGPT ads have CTRs about half the average for social media ads. However, Amir said advertisers have told him the performance of those clicks is much better than the performance of social clicks.

"The performance data is stronger than most marketers expect," he added.

Users keep chatting

But the more important signal may be what happens with ChatGPT users after these ads appear.

Users continue the conversation 73% of the time after seeing an ad, and chats continue for an average of four additional turns. That directly challenges fears that advertising would ruin the ChatGPT experience.

Instead, Similarweb data suggests most ChatGPT users are fine with ads, which appear to function more like contextual recommendations embedded inside an ongoing dialogue.

"OpenAI has a chance to take advertising to a very good place from user experience, but they can also mess it up completely," Amir said. "They're playing a good game right now. The units are not intrusive. They're very clean, very neat, small, they're in the right context."

Sign up for BI's Tech Memo newsletter here. Reach out to me via email at [email protected].