The race to master retail media has defined the past decade of digital commerce. Now, a new contender is emerging that could redraw the boundaries of how brands are discovered, considered and ultimately purchased: Generative Engine Marketing (GEM).
According to a new report from digital marketing agency Jellyfish, brands “urgently need to adopt new strategies built around Generative Engine Marketing (GEM)” as Large Language Models (LLMs) begin to govern how consumers find and choose products.
In Brands in the AI Era: Generative Engine Marketing, the agency argues that the industry is entering a phase where AI is no longer just a tool for optimisation, but an active participant in the buying journey.
“Brands have historically been made by people, for people,” says John Dawson, Vice President of Strategy at Jellyfish and co-author of the report. “Now, AI is an actor in the system: watching, recommending, choosing and even purchasing. The crucial task for brands to address is how to adapt.”
The report positions GEM as a structured response to this shift: a system designed to ensure brands are correctly understood and surfaced by AI models that increasingly act as “gatekeepers, audiences and buyers”. In practice, that means building strategies not just for human audiences, but for the algorithms that interpret, rank and recommend products across conversational search, multimodal interfaces and emerging agent-led shopping environments.
Early results suggest the approach is already delivering measurable gains. According to Jellyfish, South Korean eyewear brand Gentle Monster used LLM-driven insights to optimise its Google Performance Max campaigns ahead of the US holiday season, generating a 17% uplift in click-through rate, more than 14% growth in conversion rate and a 39% improvement in return on ad spend. Meanwhile, industrial supplier MSC Industrial Supply reported a 45% increase in revenue within 30 days of implementing AI-driven optimisations, alongside a 758% incremental ROAS.
For Dawson, the implications go far beyond incremental performance gains. “The brands that succeed in this emerging era will not simply advertise to audiences, but train the models that mediate them,” he says. “They will know their challenges and opportunities through Share of Model, manage their semantic presence and shape how AI describes their world.”
A shift in retail media logic
That idea of “training” the systems that sit between brands and consumers marks a profound shift in the logic of retail media.
For the past several years, retail media networks operated by players such as Amazon and Walmart have become the fastest-growing segment of digital advertising by monetising high-intent moments on their platforms. Brands bid for visibility on search results pages and product listings, using first-party data to target shoppers at the point of purchase.
GEM does not replace that model, it moves the battleground upstream. In an ‘ambient’ commerce environment, as described by Jellyfish, discovery increasingly happens before a shopper ever reaches a retailer’s site. AI assistants, conversational search tools and autonomous agents are beginning to curate options, narrow consideration sets and, in some cases, complete transactions on behalf of users. In that context, the question is no longer just how to win the digital shelf, but how to be included in the shortlist an AI generates in the first place.
This is where GEM starts to look like the next generation of retail media.
Instead of optimising purely for share of search or share of shelf, brands are being pushed to compete for what Jellyfish terms “Share of Model – essentially, how often and how favourably they appear within AI-generated responses. That requires a different kind of infrastructure: structured data, machine-readable content, and a consistent semantic footprint that models can interpret and trust.
“The shift really is different,” Dawson says. “For the first time, brands must build relationships not only with people but also with the AI models that surround them.”
For retailers, the implications are equally significant. Retail media networks have thrived because they control the environment in which purchase decisions are made. But as AI intermediaries begin to influence, or even bypass, those environments, retailers will need to integrate similar capabilities into their own ecosystems, embedding AI-driven discovery and recommendation into their platforms while finding new ways to monetise it.
In effect, retail media risks becoming the conversion layer rather than the discovery layer, with value shifting towards those who can shape the inputs into AI-driven decision-making.
That does not diminish its importance; if anything, it raises the stakes. Brands that are surfaced more effectively by AI are likely to arrive on retail platforms with higher intent, making downstream media spend more efficient. Conversely, those that fail to register within AI systems may find themselves excluded from consideration altogether, regardless of how much they invest at the point of sale.
GEM, then, reframes the challenge. It extends the marketer’s remit beyond campaigns and channels into the underlying signals that inform machine understanding – from product data and content to performance feedback loops and model training inputs.
“AI does not end the story of marketing, but it does open a new chapter,” Dawson says. “GEM provides a canvas on which it can be crafted.”
For an industry that has spent years refining how to influence human decision-making within retail environments, that new chapter may prove both disruptive and familiar. The principles of visibility, relevance and conversion still apply – but the audience now includes the machines that increasingly decide what consumers see.
And in that sense, Generative Engine Marketing may not just sit alongside retail media, it may define what comes next.
