Shane Ginsberg of Street Poller Media: When Fabrication Becomes Free, Proof Becomes Expensive
The economics of scarcity are usually discussed in terms of physical goods, but they apply cleanly to advertising creative, and the current moment is producing an unusually clear illustration.
Generative tools have made synthetic content cheap enough that production volume is no longer a meaningful constraint. A brand can now generate effectively unlimited variations of polished creative at negligible marginal cost, which is exactly the efficiency marketing organizations are ordinarily rewarded for capturing.
The predictable consequence is that whatever becomes abundant loses value. If every advertiser can produce visually flawless content indefinitely, visual polish stops distinguishing anyone, and the property that used to signal a serious brand becomes the cheapest thing in the category.
That leaves an obvious question about what becomes scarce instead, and Street Poller Media founder Shane Ginsberg has offered a direct answer: you cannot scale your way to trust.
His argument is that the qualities audiences actually respond to, meaning credibility and the sense that something genuinely occurred, are precisely the qualities that do not survive mass production. The efficiencies generative tools deliver come at the direct expense of the outcome the advertising was meant to produce, which makes the tradeoff considerably less attractive than the cost savings suggest.
As illustration, Ginsberg has pointed to brands like Rolex, which does not run AI-generated advertising on Instagram, instead sponsoring tennis and golf events including the PGA Tour, because that is where its audience spends attention, while maintaining a social presence built around real moments. Rolex is not a client, and the example functions purely to illustrate a broader instinct: even as companies race toward automation, brands with the most accumulated equity to protect are selecting formats that read as unmistakably human.
Ginsberg has connected this to consumer fatigue with AI content specifically, and his framing of the cause is worth attention. In his assessment audiences are tiring of AI-generated marketing not because the output is poor. Model quality has improved rapidly enough that production value no longer serves as the tell. The fatigue stems from synthetic content failing to connect the way an unscripted exchange does.
That diagnosis carries an uncomfortable implication for anyone expecting better models to resolve the problem. If audiences rejected synthetic content for looking artificial, improving fidelity would fix it. If they reject it because they suspect nothing real occurred, higher fidelity makes the situation marginally worse, since it eliminates the viewer's ability to distinguish and pushes them toward discounting everything indiscriminately.
The likely outcome is bifurcation rather than replacement. Categories where authenticity is not the operative mechanism, meaning demonstrations, explainers, animation, and anything primarily informational, should automate substantially. Categories whose entire value depends on a viewer believing a real person had a real reaction should appreciate, in proportion to how cheap fabrication becomes elsewhere.
A counterargument deserves acknowledgment. Audiences have adapted to every prior shift in advertising aesthetics, and a format reading as authentic today may read as a recognizable convention within a few years, at which point its advantage erodes on its own terms regardless of what AI does.
There is also a practical mechanism reinforcing the same direction, independent of consumer preference. Platform compliance systems have grown increasingly hostile to claims-heavy advertising, particularly in regulated categories, which means the formats reading as most obviously promotional are also the ones most likely to be rejected before reaching anyone. A person describing a personal reaction occupies different territory than a brand asserting an outcome, both in how audiences receive it and in how review systems classify it.
That convergence, between what audiences find credible and what platforms will actually distribute, narrows the range of viable creative considerably faster than either force would alone.
What appears more durable is the narrower version of the claim. As synthetic content becomes ubiquitous, verifiable evidence that something actually happened becomes a scarcer commodity, and scarce commodities appreciate. That is not a prediction about consumer psychology. It is an observation about supply.
