ChatGPT, Gemini, Google AI Mode, and AI Overviews all steer shoppers toward national chains more than independents, according to a Vaer AI study published September 29. Tom Wells, co-founder and principal consultant of Vaer AI, ran it on behalf of Lightspeed Commerce, a commerce software company. That sponsorship shapes how to read the findings. Vaer AI hands its complete methodology to those who request it instead of printing it on the page.
The dataset spans just under half a million answers to 20,000 prompts in the US and Canada. With web search off, ChatGPT and Gemini produced 200,000 answers across ten shopping categories and four cities. With live search on, ChatGPT, AI Mode, and AI Overviews added 260,000 answers and 2.5 million links. Prompts in those two parts mentioned neither store size nor the word “local.”
Without search, the models lean on training data. Large national chains were the named store in 63 to 70 percent of answers, while small or local independents surfaced in roughly one answer out of ten. When a big chain and a smaller store were paired with no labels, the bigger one won in 90 to 94 percent of ChatGPT and Gemini responses. Vaer AI sets its bands at about $1 billion or more in revenue for large, $50 million to $1 billion for mid-size, and under $50 million for small.
Concentration is steep. Roughly ten retailers take 40 to 44 percent of all recommendations, among them Walmart, Amazon, Target, Best Buy, and Home Depot. The top twelve contains zero small or local stores. National retailers took 70 to 77 percent of picks in Los Angeles and 57 to 62 percent in Montreal.
Live search narrows the gap without closing it. Depending on the model, chains were named in 46 to 58 percent of answers, versus roughly a quarter for small or local shops. Among linked stores, large and small retailers each held roughly 38 percent. On lead picks, a big chain outperformed a small one by a factor of about 2.5.
The citation data matters most for practitioners. AI Mode drew roughly 58 percent of its citations from local business listings, but only 24 percent of those became the top recommendation. ChatGPT cited large retailers in 41 percent of its sources yet recommended them first 58 percent. The first store named matched the first store cited in around 25 percent of answers. Citation did not predict recommendation.
Buying intent widens the gap. Browsing prompts gave a small shop about a third of top picks. Name a specific product and that share falls to roughly 10 percent, while a big chain’s share rises from around 40 to 60 percent. Vaer AI says AI Overviews, the surface most people see, left small or local stores out of two thirds of its shopping answers. Local stores appeared in 17 percent of its electronics answers, versus 38 to 45 percent for toys and pet supplies.
Query fan-out points the same direction. When a search query generated by ChatGPT included a store name, 96 to 98 percent were chains, large or mid-size, and 1 to 4 percent were small independents. For “where to buy X” searches, the top Google result was a social or forum post 49 percent of the time and an actual store 29 percent.
Wording mattered more than location terms. In a controlled test of 500 randomly sampled prompts across all three platforms, adding “independent” lifted small and local shops from roughly one pick in three to nearly 80 percent. The chain share of retailer sources dropped from about 44 percent with a plain prompt to as low as 9 percent, with the largest effect on Google’s platforms. “Local” and “near me” barely moved results, especially on ChatGPT. Wells attributes that to models counting a nearby chain as local.
For independent retailers, appearing in sources is not the same as being named, so measure recommendations separately from citations. Product-level specificity on store pages looks like the lever, since small shops lose ground once a shopper names an item. Local SEO teams should run “independent” prompts for their own categories and cities. Treat the figures as one vendor-funded snapshot until the methodology is public.
Vaer AI, “When you ask AI where to shop, it sends you to the big guys,” a study by Tom Wells for Lightspeed Commerce, dated September 29, 2026.
