AI Product Shelves Are Shrinking: EMARKETER's Blake Droesch Warns Brands to Rethink Visibility Strategies

Artificial intelligence platforms are presenting consumers with a narrower range of product choices than conventional retail aisles or e-commerce search engines, according to Blake Droesch, senior analyst at EMARKETER. Speaking at EMARKETER's Future of Digital Summit 2026 in New York on 27 September, Droesch told Beet.TV that this compression is creating both urgency and uncertainty for brands seeking to appear in AI-generated recommendations.

"AI is showing a smaller selection of products to consumers than they would see if they were walking down the aisles of the store or even searching on Amazon," Droesch said. "The shelf of AI can be completely random at times. So it's really a wild, wild west for retail, for brands."

Two Paths to AI Visibility

According to Droesch, large language models (LLMs) draw on brand websites, retail sites, publishers, and authoritative sources such as government agencies and medical associations when generating product recommendations. He identified two primary routes through which a brand can secure a place in that ecosystem.

"Your brand has to be well known enough by the training data that the model goes directly to your brand website, or you need to make sure that your brand has a strong presence on those retailers or publisher channels that the AI is going to for information," Droesch said. "Those are essentially the only two ways that a brand can become visible on an AI platform."

Droesch noted that large, established brands — citing Nike and Adidas in footwear — tend to dominate AI recommendations for broad queries as a predictable result of this dynamic. However, he said the picture becomes more complicated when consumers submit specific, multi-condition searches.

Specificity Raises the Stakes

Droesch argued that highly specific consumer queries — such as a shopper recovering from knee surgery, with flat feet, seeking a particular shoe size with next-week delivery — are difficult for any brand to anticipate or engineer around. In his view, consistent and intentional product positioning across every channel becomes more important as query specificity increases.

"Brands need to be very intentional about what their product is for and who it's for. Models are learning from that information and giving it to consumers in a similar way that a consumer would take in that information on their own," Droesch said.

GEO Spending Faces ROI Questions

Generative engine optimization — the practice of tailoring content and presence to improve visibility within AI-generated results — is becoming a meaningful marketing investment category, Droesch said. But he raised concerns about whether the returns justify the cost.

"GEO is getting very surgical at this point. But is your AI visibility going from 18% to 20% really going to pay off? And if that costs a lot of money on behalf of your marketing investments, is that outcome really worth it when that money could be spent on more performance based advertising?" Droesch said. "That's a legitimate question that marketers are going to really have to start grappling with once those bills come in."

Owned Websites Remain a Key Signal

Droesch pointed to EMARKETER data showing that on both ChatGPT and Google Gemini, brand websites are the most frequently cited sources in AI-generated recommendations — an owned-and-operated channel that brands already control. He also highlighted authoritative third-party endorsements and earned media in leading publications as signals that LLMs consistently weight when generating recommendations.

"If your brand's showing up there, then you have a better chance of showing up on the AI platforms," Droesch said.