"Before you take a gamble on the future of advertising in a world of AI agents, it would be wise to consider how ads have supported various breakthroughs with media and technology to date. The recent news that Perplexity is pausing advertiser onboarding sent ripples through the industry since it runs counter to the notion that ads on AI agents are an inevitability."
"The bottom line is that it's too soon to create new ad formats for AI as a media channel. Advertisers aren't ready for sponsored prompts nor any type of CPM model on AI agents. If Perplexity simply took standard search text ads and sold them on a CPC basis, it would see much better adoption. The key to success for paid search platforms has always been having enough advertiser demand to be able to show highly relevant ads for every consumer query."
"It's unlikely Perplexity has enough scale to cross that chasm. Same goes for OpenAI, for that matter. ChatGPT may have hundreds of millions of users, but it doesn't have millions of advertisers - in fact, it doesn't have a single one since it hasn't launched ads yet. But you know who does? Google. The big G has millions of advertisers and hundreds of millions of people using Gemini each month. With a flick of the switch, Google could start running search ads on AI surfaces."
Perplexity paused advertiser onboarding, demonstrating that ads on AI agents are not yet inevitable. New ad formats for agentic AI are premature because advertisers are not prepared for sponsored prompts or CPM-based models. Offering standard search text ads on a CPC basis would likely see better advertiser adoption. Paid search success depends on having sufficient advertiser demand to deliver highly relevant ads for diverse user queries. Perplexity and OpenAI lack the advertiser scale necessary, while Google and Facebook possess millions of advertisers and can enable ads across AI surfaces but are prioritizing user adoption over immediate monetization. Historical media breakpoints eventually embrace advertising.
Read at The Drum
Unable to calculate read time
Collection
[
|
...
]