The transition to AI-generated ad creative is outpacing the industry’s most aggressive forecasts, rendering WPP's recent target of 50% by 2030 a conservative baseline.
Industry conversations reveal a stark reality: public caution is masking frantic private experimentation. Small and midsized advertisers are already scaling AI creative based on early wins. As foundation models mature and adtech adopts agentic standards, 2026 will be the year this shifts from a mid-market growth hack to a mandatory enterprise lever.
AI will solve certain problems and introduce new ones.
Swapping machines for humans and the acceleration of production speed are just the two of the most obvious impacts, but these are overshadowed by the coming explosion of creative volume.
Global brands spend well over $100 billion annually on creative production, with unit costs for AI ads often being 10-100 times cheaper.
Some of that savings will be banked, but we predict that most of it will be reinvested. Because the real unlock is not cheaper creative. It is more creative.
This isn’t dynamic creative optimization (DCO), with its mechanistic swapping headlines or calls to action. It is about full executions tailored to different audiences, contexts, and geographies, still prepared ahead of time as models aren’t yet fast enough for sub-second personalization.
A campaign with millions of impressions may soon have tens of thousands of creative variants behind it, each tuned to a slightly different slice of the market. With this explosion, some ads will be gems, while many others won’t.
The deeper risks in this world are not visual glitches or slop, issues that should be technically smoothed by then. They are structural and should be concerning for any CMO.
Marketers must navigate a strange duality of AI: models simultaneously compress strategic distinctiveness while periodically spouting executional weirdness.
On one side is creative compression. Generative models are statistical engines trained on historical data. Asked for "high-performing creative," they gravitate toward past patterns. The machine sees the larger whole, causing hooks to converge, pacing to standardize, and story arcs to flatten into familiar templates. This regression to the mean compromises brands competing on distinctiveness.
Computer scientists have a term for this – semantic ablation – that describes the above-average smattering of em dashes, “delves,” and “wedges” in AI writing.
When multiple AI systems optimize across similar datasets for similar KPIs, sameness becomes the equilibrium. Originality quietly erodes because optimization pressures reward familiarity, and the models lack a nuanced understanding of exactly why something worked.
On the other side is creative divergence. Even as models compress strategic messaging, they produce a long tail of outputs that are genuinely unconventional. These include odd visual metaphors, surreal humor, or stylistic collisions no human team would deliberately sign off on. Sometimes these flashes of brilliance invent new formats. More often, they are misaligned.
As every advertiser floods the market with more variants, this long-tail weirdness raises the baseline level of stimulation in feeds. Cutting through becomes harder for everyone, shifting the arms race from making great creative to ensuring great creative can even be perceived.
And when the system is producing both a compressed “average” and a chaotic long tail, figuring out what’s actually working becomes dramatically harder.
If campaigns deploy tens of thousands of variants, performance data becomes noisier. Some assets outperform, most don’t. With limited sample sizes, random variance starts to look like insight. The “cold start learning tax” that now is accepted for larger campaigns becomes onerous with small cells. And particularly hard in campaigns with high variation.
Without structured understanding of what changed between variants – such as the timing of brand presence or the framing of the offer – learning degrades.
More creative may result in less clarity. If systems cannot attribute performance movement to specific creative decisions, they cannot compound learning. They simply keep generating, and performance suffers.
Making poor decisions, faster.
At AI’s scale, human approval for every asset will soon become impossible. It is tempting to think that because each micro-variant is seen by a relatively small cohort, brand risk shrinks proportionally.
That is a dangerous illusion, particularly for brands in regulated industries. While the reach of a single bad ad shrinks, the surface area for regulatory and compliance failure expands exponentially. A fair lending violation, a pharmaceutical misclaim, or an unsafe depiction carries legal and PR penalties whether it gets a hundred impressions or a million if the wrong person sees it. A global digital media company had to pull AI from its entire platform for months after a single machine-generated item of offensive content was mistakenly published and received a mountain of bad press.
Beyond compliance, advertising operates within other constraints: platform policies, category regulations, brand standards, on and on. Earlier in my career while working in magazines, one Japanese auto company was known to pull its ads anytime an article appeared about World War II. Our industry is full of both common-sense and similarly idiosyncratic rules.
When production was modest, governance could be centralized and manual. When production becomes effectively infinite, governance must be systematic, and compliance and brand integrity cannot be an afterthought. Guardrails must scale alongside generation, shaping outputs rather than merely blocking mistakes.
In this future, AI generation requires a new advertising operating system to evolve alongside of it, with a focus on orchestration:
Creative must move from being a series of finished artifacts to a dynamic system. That requires creative to be legible as data, not just viewable as media.
The AI creative future is incredibly powerful. It unlocks personalization at unprecedented scale and democratizes production. But scale without structure is just noise. The next chapter of advertising will not be defined by who can generate the most assets. The winners will not be the brands with the best prompts.
The winners will be the enterprises that rebuild their creative infrastructure to treat AI as a measurable, accountable system rather than a slot machine.