Shopping queries on ChatGPT doubled between January and June 2025. The platform now fields over 84 million shopping questions per week from U.S. consumers alone. Google’s AI Overviews, Perplexity and Amazon’s own Rufus assistant are all competing to become the default product advisor. When a shopper asks any of them to recommend a running shoe, a kitchen knife or a baby stroller, the same handful of large brands tend to surface. Small and mid-size retailers are largely absent from the conversation.
The reason is structural: AI engines pull product recommendations from sources with clean, well-organized data: detailed schema markup, consistent product attributes, rich review ecosystems and machine-readable pricing. Large retailers invest heavily in this infrastructure. Amazon, for instance, uses multiple LLMs to personalize product titles and descriptions for individual shoppers, cross-referencing browsing history with product attributes at a scale that no independent merchant can replicate alone. A Shopify store selling handmade ceramics or a family-owned sporting goods shop simply does not have the engineering team or the budget to build that kind of data pipeline.
This creates a compounding problem. When an AI assistant consistently recommends products from the same few brands, those brands accumulate more clicks, more reviews and more structured data, which in turn makes the AI more likely to recommend them again. Research from Liquid Web found that 65% of retail businesses are taking no steps to prepare for AI-driven product discovery. Meanwhile, the window to act is narrowing. Referral traffic from ChatGPT to major retailer sites grew so quickly in 2025 that the platform now represents over 8% of Amazon’s weekly search volume, up from less than 1% a year earlier.
The workable path forward is infrastructure that translates existing product catalogs into content AI systems can parse, recommend and link back to, without asking every small retailer to hire an AI engineering team. That means generating structured, semantically rich storefronts that serve both human shoppers and the AI crawlers that increasingly send those shoppers to their doors. The prerequisite layer is already working in production. Evan Alexander Grooming, a men’s grooming brand with 179,000 YouTube subscribers, had strong traffic and content authority but served every visitor the same static storefront. After deploying dynamic storefronts that personalize navigation, product highlights and on-page Q&A to each visitor’s goals, they saw a 2.09x conversion lift and a 207% increase in revenue per visitor. Measuring AI visibility directly is the next step; the storefront foundation it depends on is proven.
This matters beyond individual business outcomes. Small retailers account for a significant share of U.S. employment and local tax revenue. NBER research has shown that ecommerce expansion already reduces retail employment in affected counties by roughly 2.9%, with the sharpest losses among smaller and newer stores. If AI-driven shopping accelerates that consolidation, the consequences reach into communities that depend on independent retail for jobs, local identity and economic circulation. A 2025 Shopify report noted that nearly 90% of retailers are now actively using or assessing AI, but for most small merchants, “assessing” still means watching from the sidelines while the market shifts beneath them.
The gap between assessing and deploying is where most small merchants stall. Enterprise retailers close it with headcount: a data team to structure the catalog, an engineering team to integrate personalization, a budget to run experiments. A ten-person Shopify store cannot. The fix has to be architectural. Intent detection, behavioral personalization, and dynamic content at scale were never out of reach because they were inherently complex to use. They were out of reach because nobody had built them for a small merchant. With the right abstraction layer, the merchant simply connects their catalog, deployment drops from months to days and results show up within weeks.
Brands that build AI-readable product data now will earn visibility that compounds over time while those that wait will find catching up steadily more expensive and, past a certain point, no longer possible. Whether AI-driven commerce becomes another consolidation engine depends on how quickly the infrastructure to keep independent retail in the conversation actually gets built.
Saran Kumar Krishnasamy is the Co-founder and CTO of Gigit.ai, where he’s pioneering Generative Engine Optimization (GEO) to help ecommerce brands maintain visibility as consumers shift from traditional search to AI-powered product discovery. He previously scaled the AI team from 5 to 50 and built real-time NLP systems processing petabyte-scale data for global event detection as Staff AI Engineer at Dataminr. Saran has over a decade of experience building production AI systems at companies including Visa and PayPal, with expertise spanning large language models, agentic AI systems and the intersection of AI and commerce.
