"Since the ad push started last month, Digiday has contacted more than 23 agencies. Most said what they knew about the fledgling ads business came via leaks, gossip or reporting, not OpenAI. The few who were aware of the pitch said they heard it from clients as OpenAI has been approaching brands directly. Adthena, for example, whose 400-strong client roster includes a number of household names, such as The Home Depot, Dell and Inspire Brands, has noticed the caliber of brands OpenAI has targeted."
"Those talks have largely centered on OpenAI asking marketers to commit a share of their total search budgets, according to two ad execs familiar with the meetings. In some of those instances, talks started around a $250,000 minimum commitment. In others, agencies said the number could be lower or higher depending on how negotiations shake out. Whatever those figures net out to, the dollars will fund a test period in the U.S. lasting several weeks, beginning on February 6. During that period, marketers will buy sponsored placements at the bottom of answers on the free and Go tiers of ChatGPT at $60 per 1000 views. That rate is comparable to targeted streaming and premium TV inventory like NFL games, media buyers say, and well above Meta's typical CPMs, which often come in under $20."
Agency networks and marketers learned about ChatGPT advertising largely through leaks, gossip and direct client outreach rather than official OpenAI communications. OpenAI has been approaching the largest brands directly, with even medium-big clients often not contacted. Initial talks focused on marketers committing a share of their total search budgets, with some discussions starting around a $250,000 minimum commitment and negotiable terms. The committed dollars will fund a U.S. test beginning February 6, during which marketers can buy sponsored placements at the bottom of ChatGPT answers for $60 per 1,000 views. The rate compares with targeted streaming and premium TV CPMs and exceeds many social-platform CPMs.
Read at Digiday
Unable to calculate read time
Collection
[
|
...
]