Can We Mitigate the Economic Risks of AI in Digital Advertising?

By Kenneth Rona, Ph.D., Chief AI Officer, JWX

The arrival of AI to the digital advertising landscape has brought a lot of excitement, as well as a lot of smoke and mirrors. As someone who is on record as cautiously pessimistic about, I believe we haven’t taken enough caution in assessing the risks that come with AI adoption.

By all means, AI should work in advertising, especially when it comes to improving campaign performance and uncovering new areas of efficiency for both media buyers and sellers. For as much as the industry hypes the upside (better performance, lower operating costs, and more automation), we neglect the potential downside.

This includes the economic risks, including who benefits, who loses leverage, and who gets squeezed. AI will not enter a neutral market; it will land in an industry already defined by concentration, asymmetry, and unequal bargaining power. The introduction of AI, without any kind of safeguards, will result in rapid deskilling, dependence on large intermediaries, IP and compensation disputes, and creative homogenization that reshape who captures value.

Let’s examine these risks more closely.

Deskilling and Strategic Drift

One of the quieter risks AI brings to the buy side is deskilling. As planning, optimization, targeting, reporting, and creative iteration get automated, marketers may lose some of the very capabilities they need to challenge the system.

If this continues, buyers may find themselves in the strange position of running more optimized campaigns with less human understanding of why they work — or whether they are working in the way that matters. AI can improve execution while hollowing out judgment. That is a dangerous trade for an industry that already struggles to distinguish easy measurement from meaningful measurement.

There is also a content and IP problem hiding inside the seller story. Publishers and creators increasingly risk being used as raw material twice: first as places to collect audiences, and second as content inputs for model training, summarization, or synthetic creative systems.

That uncertainty matters. If AI lowers the cost of producing “good enough” content while muddying ownership and compensation, it will cheapen the market even as it expands supply. Publishers may find themselves competing not only for attention, but against derivative systems partially built on their own work.

Homogenization and Lack of Innovation

AI is excellent at producing outputs that resemble what has already worked. That is useful in performance marketing and dangerous in brand building.

If every advertiser uses similar models trained on similar priors to optimize similar messages against similar metrics, then campaigns may become more efficient and less distinctive at the same time. Buyers may get cheaper iteration and faster testing, but at the cost of originality, surprise, and real differentiation. That is good for short-term throughput. It may be terrible for long-term brand memory.

For sellers, the downside is a market increasingly flooded with competent, cheap, forgettable content. In advertising, that means more sameness: more convergent messaging, more templated creative, more optimization toward established winners, and less genuine differentiation.

The likely result is a market full of content that is efficient but not especially memorable. Buyers may get efficiency. Sellers may get a race to the middle. Audiences may get bored, but that boredom will probably just trigger another round of investment and innovation. The system will respond to the problem by doubling down on the thing that helped create it.

Market Structure and Dependency

Then there is market structure. AI in advertising will not be deployed on a level playing field. Buyers that depend on a small number of powerful firms for optimization, measurement, creative generation, data handling, and infrastructure may gain automation inside workflows while losing autonomy at the market level.

That means the buy side may become more operationally efficient while also becoming more strategically dependent. AI may reduce toil while increasing captivity.

The same concentration pressure applies to sellers, and in some ways more sharply. For publishers and supply-side players, AI may not just be another tool layer. It may become another dependency layer. Sellers may gain optimization, packaging, and monetization support, but they may also become more reliant on firms that sit upstream of them in infrastructure, data, and model access.

That is not just a commercial issue. It is a bargaining-power issue. The big players will have the resources to build, buy, or negotiate their way into advantage. Smaller publishers will not.

Net results

The cautiously pessimistic view is not that AI will fail in digital advertising. It is that it will succeed in the wrong way, making the industry less competitive and less skilled.

Some lives will get easier. Some jobs will get thinner. Some companies will become much more productive. Others will become more replaceable. The big players will win. Small publishers will suffer.

The real question is not whether AI will make advertising more efficient. It almost certainly will. The real question is whether the industry can prevent “better performance” from becoming a euphemism for worse markets, weaker trust, and bigger external costs.