ChatGPT Ads: Limited Reach and Uncertain ROI

Over 100,000 brands have tried ChatGPT ads. With limited reach, imperfect measurement, and uneven results compared to Google, this channel is still searching for its place in media plans. OpenAI is rolling out tracking and bidding tools, while some advertisers are beginning to structure more repeatable approaches.

Reach and Targeting Hinder Scaling

Several factors are slowing the progress of OpenAI's advertising activity: the free user base of ChatGPT remains limited, few conversations truly lend themselves to ad insertion, and there is a limit to the number of ads users tolerate before leaving the platform. Dave Dugan, who oversees global advertising solutions at OpenAI, stated in June that the reach and quantity of ads served have progressed alongside the increase in user experience trust. Don McGuire, Chief Marketing Officer of Qualcomm, believes it remains complicated for language models to reach an audience, and targeting only becomes feasible on very broad groups, raising doubts about profitability.

Better Measurement: Pixel and Conversion Bidding Arrive

According to Ashley Fletcher, CMO of Adthena, the question of evaluating advertising effectiveness remains a challenge for marketing professionals. OpenAI has implemented a new tracking pixel and introduced a conversion bidding system that allows for setting a target cost per acquisition or per click, aiming to reduce uncertainty. Fletcher believes these tools will provide more relevant data for organizing the next steps, but they may also highlight that ChatGPT ads primarily intervene at the beginning of the customer journey, far from the quick conversions typically observed on other platforms.

Performance: Gap with Search, but Early Positive Cases

According to a working paper by Maximilian Kaiser and Christian Schulze covering 102 advertiser sites, ChatGPT ads generate $0.89 in revenue per session in the U.S., compared to $2.45 for Google search ads. The observed performance gap does not guarantee a simple catch-up over time. Between July and September, Adthena reports that 52% of the 30 major brands studied did not recoup their spending on ChatGPT, with 26% matching or exceeding Google, while the rest showed mixed results. OpenAI highlights a case study published in October: at Evernote, the conversion rate from trial to paid subscription via ChatGPT reportedly reached about 10% to 20% above the historical average on paid channels earlier this year.

How Advertisers Are Organizing to Leverage the Channel

More than 100,000 advertisers have already tested ChatGPT ads, but their sustainable adoption will depend on the ability to demonstrate an impact on key business metrics. Chris Koehler, CMO of Twilio, mentions trials, while Thomas Ranese, CMO of Intuit, speaks of ongoing experiments. Some companies are structuring their approach: at AdRoll, Hans Fischmann explains that each ChatGPT ad is tagged, and this data is used to guide overall media buying in a full-funnel logic. Natural Intelligence, which manages Top10.com and BestMoney.com, has shifted its ChatGPT ads from its test budget to a more repeatable growth budget. Jonathan Vardi, head of strategy, indicates that transforming pain points into targeted ads and accurately conveying conversion data to the platform are essential for training the advertising algorithm. According to him, limiting oneself to tracking views or clicks is no longer sufficient to ensure a satisfactory return on investment, especially in a context of increased competition. Industrially, OpenAI started serving ads about nine months ago, has enriched its offering with new formats, partnerships, and tools at a steady pace, and reached an annualized revenue of one billion dollars this summer.