AI Is the New Search. Here’s How Retailers Win the Answer
Shoppers now start with AI, which picks the best fit, not the biggest retailer. That gives the edge to those that articulate their distinct value.
Par Aaron Cheris, Mikey Vu, Darrell Rigby, Naomi Alford, et Stephanie Koszyk
Retail Holiday Newsletter
AI Is the New Search. Here’s How Retailers Win the Answer
AI Is the New Search. Here’s How Retailers Win the Answer
Shoppers now start with AI, which picks the best fit, not the biggest retailer. That gives the edge to those that articulate their distinct value.
Par Aaron Cheris, Mikey Vu, Darrell Rigby, Naomi Alford, et Stephanie Koszyk
First published on novembre 21, 2026
min
Etude
AI Is the New Search. Here’s How Retailers Win the Answer
en
En Bref
About a third of consumers in the US and Europe use AI assistants for shopping, so retailers absent from the answers will cede visibility, traffic, and profits.
AI often overindexes on category specialists and insurgents over generalists and market share leaders.
Making the list requires a clear value proposition, an adaptable generative engine optimization (GEO) strategy, and a focus on winning the shopper’s occasion.
Leaders will also build their assistants on proprietary data, using personalization, brand trust, and category expertise to become the place shoppers start.
AI is becoming the new search. It’s reshaping the top of the shopping funnel.
No retailer would choose to be invisible on Google. Most have spent years honing search engine optimization so shoppers seeking a specific item find them first. Now, ChatGPT alone serves 1.2 billion weekly users, and an estimated 15% to 20% of their prompts are shopping-related, up from 9% in 2025. A Bain Consumer Lab global survey shows that 33% of consumers in the US and 36% in Europe say they use AI assistants for some part of the shopping journey as of November 2026, up from 19% in the US and 8% in Europe in September 2025.
These shoppers are asking AI assistants to plan end-to-end occasions: “Help me host a holiday dinner for eight.” The assistant recommends a broad basket of goods and services from a handful of retailers, a far shorter list than Google’s pages of results. Earning a spot takes GEO, which shapes what AI models know, trust, and say about products and retailers. Yet many retailers are leaving GEO to chance.
Yes, AI assistants such as ChatGPT and Claude currently account for less than 1% of retailers’ directly linked web traffic, according to our analysis of estimates from Similarweb, a digital market intelligence company. But click-through data undercounts their influence. Many of these shoppers use AI to research and summarize, then go to retailers’ sites on their own. And that AI referral traffic, though small, converts 90% better than non-AI traffic.
The battleground is moving from the search results page to the AI answer itself. The fight is over the ever-narrowing path to the visit, online and in stores. Retailers absent from AI responses will cede visibility, awareness, traffic, and profits.
Leading retailers will compete for traffic on three new but rapidly growing fronts: third-party (3P) assistants, first-party (1P) assistants, and customers’ personal AI agents, such as Meta's Muse. Showing up on 3P assistants is table stakes. Winners will pair a 3P strategy with their 1P assistant, giving shoppers reasons to start with them through personalized recommendations, category expertise, and loyalty benefits that outside assistants can’t match.
The name of the 3P game is AI visibility, and market share doesn’t automatically guarantee it. We analyzed more than 4,400 shopping prompts across four retail verticals, testing more than 300,000 unique runs across the most used large language models (LLMs), powered by Profound data. For many categories, specialists appear in AI answers more often than generalist retailers, relative to their market share (see Figure 1).
Figure 1
Retail market share doesn’t always translate into AI visibility
Grocery, a lower-consideration category with a broader retailer base, is more fragmented than specialty categories. Mentions depend on retailers’ strength in the purchase dimensions shoppers seek. For instance, Kroger appears more often for prompts about promotion and loyalty, whereas Whole Foods outperforms on dietary needs and product selection.
Prompts vary by consumer intent and purchase complexity, as our shopping mission framework shows. Directed or spec-defined prompts, such as “the cheapest four-pack of AA batteries by tomorrow,” are won by item. Retailers have historically chosen the items they want to be known for and optimized for targeted “where to buy” searches on precise dimensions such as price, quality, or service. Now, broader missions, such as “help me throw a holiday dinner party,” put the whole basket in play. Retailers must also win these open-ended “what to buy” prompts, where no one owns the answer yet.
Early AI adoption alone won’t turn a weak retailer into a strong one. A superior, well-communicated core value proposition should still win the purchase. Leading retailers will continually ask, “Are we winning for the things we want to be known for?” If not, either the value proposition isn’t holding up under full transparency, or it isn’t explained well enough. Three actions can help:
Make your advantages compelling or compete on price and speed alone. An LLM optimizes for what it can find, typically assortment and price. In the future, assistants could default to the cheapest possible basket across multiple retailers, unless they have better information about each retailer’s value proposition. Think of the 3P assistant as the consumer’s delegate. Retailers wouldn’t hand shoppers a catalog and price list without the brand behind it, so they shouldn’t hand one to the shopper's assistant either. The more complex a retailer’s offering—such as add-on services, a rewards program, or quality differences between seemingly similar products—the less likely it is to surface in the AI answer. Successful retailers spell out who they’re best for, what exclusive products they have, and which occasions they serve.
Build a responsive GEO strategy.The citations LLMs draw on are highly fragmented (see Figure 2). In our grocery analysis, the top 10 sources accounted for roughly 30% of all citations across ChatGPT, Claude, and Google AI and Gemini. Even so, earned media currently leads for most LLMs. But that mix will keep shifting: A recent ChatGPT update, for example, sharply cut social media citations in favor of retailer websites and earned media. Winners will build structured, readable, accurate site content and back it with credible third-party media coverage. Because AI answers shift for reasons inside and outside a retailer’s control—a platform updates its model, a competitor changes its offer, a PR campaign lifts media mentions—a solid GEO strategy never stops testing, learning, and adapting.
Figure 2
Third-party AI assistants’ cited sources are highly fragmented, but mostly favor earned media
Go beyond the item to win the basket.Identify the occasions that matter most to customers, such as “quick dinner tonight” for a grocer or “travel capsule wardrobe” for an apparel player. Then organize product and content data accordingly and support that story through media coverage, creators, and communities. On 3P assistants, retailers must rely on the LLM to suggest complementary products, such as speakers and cables for a new TV. Some are responding with proprietary apps in ChatGPT to personalize recommendations and integrate loyalty programs. Retailers can also sponsor ads or set up branded agents, such as ChatGPT's Sponsored Agent pilot or Google's Business Agents, which could suggest speakers to go with the TV.
1P: A strategic imperative
The remaining 99% of traffic to retailers’ websites that still comes from traditional sources, including search and direct visits, is at risk. Retailers with high e-commerce penetration or in high-consideration categories, such as automotive or furniture, need a 1P AI assistant to protect it. Yet only 27 of the top 100 US retailers have one, and the quality and usefulness of those that do vary widely.
Of the top 100 US retailers
have a 1P AI assistant
Retailers that rely solely on 3P assistants risk losing the primary customer relationship and the proprietary data that comes with it, both of which are hard to win back. Restaurants and hotels learned that lesson from food delivery and travel aggregators. Hand over the keys, and someone else decides where the car goes and charges for the ride.
3P visibility is a necessity, but 1P is the engine of a strong AI strategy. A Bain survey found consumers still trust retailers’ 1P assistants nearly twice as much as 3P ones. An owned assistant gives shoppers a reason to start with the retailer instead of a generalist chatbot that could steer them elsewhere. Rich customer data and personalization opportunities make the basket easier to win. Apps with assistants compound the advantage with relevant, personalized experiences and loyalty benefits for logged-in shoppers.
Retailers are already reporting results from AI assistant users compared with nonusers. Walmart says Sparky users have up to 40% higher average order value; Amazon says Alexa for Shopping users have about 40% higher average order value. For Brazilian retailer Magazine Luiza, shoppers using its Lu assistant via WhatsApp convert at three times the rate of its traditional app users.
Four steps can help retailers build 1P assistants quickly and effectively:
Make the assistant a brand ambassador.A great 1P assistant draws on merchant and store associate expertise that a generalist chatbot can’t match, while building on hard-earned customer trust through a recognizable persona and voice. Best Buy’s Ask Blue, launched in time for the holidays, replicates the expertise and support of its blue-polo-wearing associates. Magazine Luiza’s Lu is friendly, warm, and fun to chat with on WhatsApp. Lu earns an 85-point Net Promoter Score℠, and 75% of payments happen on WhatsApp.
Give shoppers a reason to visit.AnAI assistant is the best channel to articulate a retailer’s differentiated offerings, such as exclusive assortment, low prices, or add-on services. Carrefour’s Hopla helps shoppers eat healthier with product swaps based on dietary preferences and analysis of cart and meal photos for more nutritious recommendations.
Own the occasion. Retailers know their shoppers’ behaviors and their products’ return, repeat, and cross-purchase rates better than any 3P assistant. They can train assistants to build the basket with follow-up prompts and personalized add-ons, going beyond single products to capture the spirit of the ask.
Protect retail media revenue.Leading retailers keep ads in their assistants consistent and relevant to maintain trust. As AI routes shoppers directly to product detail pages, retailers can protect retail media profit pools by producing higher-value ad units from rich intent data and by expanding in-store and off-site media.
How to win this holiday season and beyond
There’s a new twist: the recent launch of personal agents, such as OpenAI’s dots and Meta’s Muse. In the six weeks after Muse launched, more than 10 million people in the US and Canada set up agents. Shopping is the No. 1 use case, with purchases averaging more than $200. If early trends hold, consumer adoption is set to skyrocket in 2027.
The year is almost over, but retailers can’t afford to coast through the holiday season. Here’s a checklist leaders can complete before the New Year’s countdown begins:
Run a prompt audit. Test your 3P visibility for the items and occasions your customers care about most, considering their intent and how they actually shop, before Black Friday and the final holiday rush.
Close GEO gaps to ensure your value proposition comes through. Update your most important pages for the occasions you want to own and the promotions you want shoppers to see, using your customers’ language to describe you. In addition, start thinking about how personal agents, which prompt and behave differently from human customers, will change your strategy.
Check on your 1P assistant roadmap.If you don’t have a launch plan for early 2027, consider a white-label solution or an outsourced build to get there faster.
Get your most important data ready.Assess whether your consumer and product data is ready for hyperpersonalization in your 1P assistant. If not, move quickly on the highest-value data first.
Determine your agent-to-agent strategy.Assign an owner and put pen to paper before the holidays. Take a stance on how open your 1P assistant will be to outside agents.