4 Comments
Oct 29, 2022·edited Oct 29, 2022Liked by Ryan Khurana

Some great points. Thanks for this article. I’ve been in creative automation for brands for a number of years.

I will say that - aside from the types of “wow isn’t this AI stuff neat!?” type of campaigns you mention - the capabilities are already FAR ahead of where a lot of brands have been willing to test.

The idea of contextually adapting creative to align with content has been around in the form of Dynamic Creative Optimization for a number of years - dynamically inserting contextually relevant copy, images or CTA via macros into an ad based on pre-defined rules - but brands have been slow to adopt it and reluctant to hand over real decisioning to the algos even when we have had that capability for years.

I remember presenting an ambitious DCO campaign that would leave limited decisioning to the machine to a well known US automotive manufacturer a few years back. The CMO immediately shot it down and opted for a narrowly defined rule set because “What if the machine creates something that isn’t ‘brand safe’? We can’t take that risk.” Mind you this was just remixing existing creative elements in a database that the brand had already approved!

As you point out a lot of the early adoption will be among SMBs and performance marketers who don’t have limitless creative budgets. It’s going to be a fun industry!

Expand full comment

This is the best post I've read on the topic – very forward thinking!

I'm coming at generative advertising from the opposite angle: generative AI services that want to monetize through ads have a problem: when you don't know what your users will generate, neither do advertisers. This breaks targeting and threatens a brand safety nightmare.

You actually need something much more like search ads on generative AI sites – but Google doesn't have a 3rd party search-display ad product.

Not every AI product can use SaaS pricing to cover their GPU costs. A bunch of generative companies are about to realize advertising is broken for them.

Expand full comment
author

Thanks!

And I'm not sure if that's the case for established products. You have a history of a user's generations similar to a history of tweets that would let you understand what they're likely to make. I think getting from a stream of generations to predicting what you will like to see next is increasingly becoming a trivial task.

Expand full comment

Good point - the data to model preferences is definitely there. I wonder how many publishers will choose to do their own modeling vs hand it off to the ad-tech stack. Sending user prefs to advertisers in a form they can use is still hard. Usually involves partnering with a data management platform to match cookies and send the user data to the ad auction.

Then there's targeting real-time intent on the current prompt. I think the coolest generative search ads may be collaborations between search platforms’ AI and brands’ AIs. Here’s how I imagine AI search integrating an REI ad: https://www.stratos.blue/images/gear-ad-embed.png

Expand full comment