The inflationary impact of AI-generated ad creative

Last week, Meta unveiled its AI Sandbox: a suite of generative AI tools designed to assist advertisers with creative production. These tools work across text and image assets and functionally serve to facilitate the generation of advertising creative variants. Meta’s CTO had hinted at the imminent availability of such tools last month in an interview with Nikkei, so their announcement was not altogether surprising. Google, similarly, revealed in a presentation to a select group of advertisers last month — titled “AI-powered ads 2023” — that it will begin offering generative AI tools for the purposes of creative production this year.

I state in Abandoning intuition: using Generative AI for advertising creative, published in December of last year, that, “of the various enterprise use cases served by generative AI, none is perhaps more immediately actionable than advertising creative production.” Anecdotally, I have reviewed more than one dozen pitches for generative AI tools for advertising creative production in the past six months. The use case, as I describe in Abandoning Intuition, is clear: generative AI can not only help marketing teams reduce the cost of creative production but also increase the scope of their experimentation, relieving teams of the burden of deriving new concepts and potentially unlocking novel aesthetic or product marketing themes that are non-obvious (hence the abandonment of intuition).

But, similar to offshoring or process automation, the primary value proposition of these tools is that they reduce cost. One dominant trend in digital advertising for the past several years has been mass-volume creative production and experimentation in pursuit of the efficient pairing of ad creative with relevant audiences. I discuss this dynamic in Mobile ad creative: how to produce and deploy advertising creative at scale, which was published in September 2019. From that piece:

A first thing to clarify in starting the post is that ad creative is not suddenly important because Facebook and Google automate campaign management: ad creative production has always been critically important to mobile marketing teams. What has changed now is that Google and Facebook can target and segment groups of users so much more specifically and efficiently with their automated optimization schemes — AEO and VO on Facebook and Google’s UAC — that testing very many ad creative variants against every segment of potential audience is now possible.

The ability of ad platforms to optimize targeting on the basis of user-level behavioral profiles as described in that piece has been disrupted by platform privacy policies, such as Apple’s App Tracking Transparency (ATT), and regulation. Ad platforms have adapted to these restrictions by utilizing any number of the four means of increasing advertising revenue that I articulate in Unpacking Meta’s pivot to an open graph and short-form video. Those are:

1. Increase “ad load,” or the ratio of ads shown to each user per session relative to organic content;
2. Increase reach, or the number of users that engage with a product and thus are exposed to ads;
3. Increase the value generated by ads through higher-quality formats or better targeting, which improves the general price paid for ad inventory through increased bids from advertisers in the ad auction;
4. Increase time spent on site, which provides more opportunities for ads to be served.

This playbook has been activated by ad platforms through applying machine-learning automation, characterized as AI, to on-site, first-party data in two ways:

1. Personalizing short-form video content (eg. Meta Reels, YouTube Shorts) to increase user time spent on site and thus ad exposures (method #4 above);
2. Rapid, automated audience targeting and creative experimentation (eg. Meta Advantage+, Google PMax) to minimize campaign management inefficiencies (method #3 above).

Applying machine learning-driven automation to creative production accelerates this trajectory to its natural logical conclusion, which is end-to-end marketing automation using on-site performance metrics (such as clicks) as inputs. I make the case in Exoskeletons, not cyborgs, and in Meta’s ad spend glitch and the risks of marketing automation, that advertisers should be wary of relinquishing to platforms the control of their brand identity and, for many companies, their foremost revenue driver and largest single line-item expense. That said, the allure of generative AI for creative production is hard to deny: Meta, Google, and other ad platforms already possess performance data for substantially every single digital advertiser. If that storehouse of data can be brought to bear on creative production through a cross-advertiser data co-op — much in the same way it was utilized for event-level campaign targeting — then advertisers that don’t participate are put at a competitive disadvantage.

Value is generated in digital advertising through targeting; the large ad platforms are best positioned to capture this value through the integration of generative AI tools with their own streams of traffic. While AI-empowered creative production and campaign optimization aren’t paired and fully automated into platform-specific workflows yet, it’s only a matter of time before they will be. Advertiser-specific creative production will be supported automatically through native platform tools that build and deploy ads in real-time on the basis of performance feedback. The more money spent, the more feedback gathered, and the better that advertiser’s ads will perform.

Generative AI is deflationary for content production but is inflationary for distribution. Generative AI will see the production costs of increasingly-complex forms of content (eg. video) approach zero; these tools will instigate an immense expansion in the volume of each content format that they perfect. The first photograph to feature a human being was taken by Louis Daguerre, inventor of the daguerreotype process, in Paris in 1838. According to The Guardian, as a result of widespread smartphone ownership, 1TN photographs were taken in 2014, representing more than a quarter of all existing photographs taken up until that point. Statistics like this will echo across text, animated and photorealistic video production, audio, etc. in synthetic form as a result of generative AI.

And as content proliferates through generative AI tools, the challenge of capturing potential customer attention becomes more acute — necessitating an increased reliance on advertising. This is inflationary: the corpus of content will grow at a much more rapid pace than the human birth rate. Organic discovery becomes ineffective as content mushrooms; this dynamic gave birth to the search ads mechanism in the first place. Generative AI will similarly create competitive friction for the discovery of all forms of content.

And as a corollary: while the mass generation of conversion-optimized ad creatives by platforms decreases the cost of advertising asset production, it will result in increased competition for scarce ad impressions. The compression of creative iteration timelines will simply put upward price pressure on auction mechanics as real-time creative optimization automates away the last non-monetary lever for improving campaign performance: creative.

If every advertiser is operating at maximum creative efficiency through the use of shared, platform-centric creative production tools, and creative iteration takes place at the blinding speed unlocked by generative AI, then the only recourse any advertiser will have to improve performance in auctions is through increased bids.