Adapting to Retail Consumer Behavior

Retailers shift from Google to AI agents – what this means for FMCG brands A silent shift is underway in digital commerce — and FMCG brands should take note. In August 2025, ChatGPT drove 20% of referral traffic to Walmart and Etsy Shop, with Target at ~15% and eBay at 10%. Just a month earlier, these numbers were significantly lower. While referral traffic is still under 5% of total visits, the growth velocity is clear. Consumers are replacing search with conversation. Instead of using Google, users now ask ChatGPT: - “Which toothpaste is best for sensitive teeth?” - “Top healthy snacks for kids?” - “Why is Swiss Cheese so good and where can I buy it?” - “Best laundry detergent for cold wash?” This behavioral shift matters. AI agents filter and surface product recommendations based on trust, brand recognition, and relevance — not just ad spend. For FMCG producers, the implications are clear: – Visibility is no longer guaranteed by shelf space or SEO. – If your brand isn’t part of AI agents’ product surfaces, you’re invisible. – Retailer data access policies now shape your discoverability. Retailers like Walmart (420 million SKUs) and Target are gaining exposure by remaining open to AI crawlers. Amazon, however, has blocked many bots — causing its ChatGPT-driven traffic to drop 18% in August. This evolving ecosystem affects how FMCG brands are discovered, recommended, and ultimately purchased. And unlike paid search, where placement is auctioned, AI-driven recommendation engines operate in more opaque, model-based hierarchies. Key facts: – 2.5 billion daily ChatGPT prompts – ~50 million daily shopping-related queries – 60% of US shoppers have used genAI for shopping (Omnisend, Aug 2025) As OpenAI and others move toward affiliate fees and embedded checkout, FMCG brands must act now — ensuring their products are correctly indexed, accurately represented, and promoted within retailer ecosystems that are embracing AI traffic. The next shelf is conversational. And it's already stocked. #retail #ecommerce #fmcg #omnichannel #ai #chatgpt #openai #shoppingagents #digitalcommerce #referraltraffic #amazon #walmart #etsy #target #ebay #rufus #retailtech #consumertrends #searchvschat #affiliate #onlineshopping #generativeai #shoppingbots #conversion #usa #northamerica #martech #digitalmarketing #adtech #aiincommerce #futureofshopping #platformeconomy #brandvisibility #fmcgmarketing

A week after Shoptalk Europe 2025, and after a week of reflection on all the content and conversations, one major theme has emerged for me: the importance of digital influence. In the Shoptalk European Retail Zeitgeist this year I wrote “whilst the point of purchase still remains physical in the majority of cases, for the majority of categories and many retailers and brands the points of influence, of critical demand creation, are increasingly digital”. I wanted to dig into this further. On the Keynote stage, in my interview with Marc Carena of Mars Wrigley, he captured this as the need to drive both “mental and physical availability”. Many consumer brands are well-oiled machines at driving physical availability (points of distribution), so what does driving mental availability mean in an age of digital influencing? Three main implications emerged at the show; 💲Media Investment Allocation During my interview with Jordi Bosch Argilagós of Nestlé, he shared that over 70% of Nestlé’s media investment is already digital, as “data unlocks personalization and personalization drives better shopping experiences and better ROI”. Marc Carena shared that Mars Snacking has more than doubled media investment over the last five years, and switched the mix heavily in favour of digital (to a similar level as Nestle shared above). Going further, Marc outlined the Mars strategy to increase reach and engagement by investing in “Earned” and “Shared” media channels, over “Paid” media. 👍 Social Commerce Digital media investment is focused on three areas; Search, Social and Retail Media. And whilst Retail Media is the fastest growing, Social has overtaken Search to be largest category in some European markets. Social platforms are also growing in importance as commerce platforms, and Vladimir Hanzlik shared new EMARKETER data highlighting that nearly 20% of shoppers in the five largest European countries made a purchase directly from social media last year (slide below). In his Track Keynote, Mark Elkins shared L'Oréal’s learnings on social commerce, especially on TikTok and Douyin. He stressed the importance of authenticity, of providing advice, of creating emotional experiences whilst also offering convenience, and of reviews. 📏 Measure the Impact We should expect the importance of digital influencing to grow. Mike Black shared Profitero+’s latest research illustrating how younger shoppers over-index in being influenced digitally (slide below). Measuring the impact of digital influence, rather than ecommerce sales, remains hard and continues to be a barrier to unlocking investment for some. However, advances are being made in this space, and Prasanna Kumar 彭天乐 shared the work he is doing with Danson Huang at Diageo to quantify the level of digital influence, and help “bring the digital commerce agenda onto the centre table” (slide below). 🚨 So, are you set-up to drive both physical and mental availability in your organisation? 

AI is accelerating structural shifts across the retail sector, and this article provides a clear view of how the landscape is evolving — from changing consumer discovery journeys to the growing influence of AI-driven recommendation systems. The piece highlights several areas that merit close attention from retailers and brands: 💳 How purchasing decisions are increasingly shaped by AI intermediaries, not just traditional marketing and search. 📊 The need for robust data foundations to ensure products are accurately represented in AI-driven environments. 🤖 The operational and regulatory considerations around autonomous shopping agents and automated decision-making. PwC UK’s Emma Ford and Anna Bancroft help frame the practical challenges and opportunities we’re discussing every day with our clients. 🔗 in the comments #Retail #AI #DigitalTransformation #PwC #ConsumerMarkets #BusinessStrategy

McKinsey & Company published a report on shopping in the age of AI. Recommended read. I wanted to extend this to the retail media implications. 🛒 The headline: AI is bifurcating the shopping journey into convenience-driven trips (delegate to an agent) and discovery-driven trips (go for the experience). Store visits become LESS frequent but MORE valuable. That tracks. But here's what the report doesn't address: what happens to retail media when the convenience trip gets intermediated by an agent? Retail media is one of the fastest-growing ad categories in our industry. Its entire value proposition rests on the assumption that consumers visit retailer-owned properties, where their behavior can be observed and monetized. Agentic commerce challenges that assumption in digital. If an AI agent assembles the basket, compares alternatives, and executes the purchase upstream, the retailer's site becomes a fulfillment station and not a media surface. The pace of this shift is the part I think most people are getting wrong. I keep coming back to online banking as the analogy. Wells Fargo launched online banking in 1995. It took until roughly 2010-2015 for it to become majority behavior. That's a 15-20 year curve for technology that was fully functional by 2000. 👉 Protocol readiness is not the same as consumer adoption. Let me say that again for the people in the back. PROTOCOL READINESS IS NOT THE SAME AS CONSUMER ADOPTION. Trust compounds slowly, through repeated low-stakes successes. My adorable mom is not delegating her grocery order to an agent next year. However, and this is the part the cautious camp misses, the infrastructure being built right now (ACP, UCP, Agent Pay, what platforms like Mirakl are doing on the inventory side) is what determines who is positioned when behavior catches up. The aspirational (searching for my new Porsche Cars North America) and the actualized (buying a Honda) have far different data value so don't discount AI for eventually getting in on that sweet sweet transaction. For retail media specifically, this means the next few years are about building toward agent-readability: structured product data, machine-readable pricing and fulfillment, and partnerships with the payment-data layer that will increasingly inform what agents recommend and how those agents are secured. The networks that do this work now will be in da club and the ones that don't may be bypassed. That said, bypass isn't the only outcome. Retail media can and will adapt: advertising upstream in agent recommendations, ranking influence via structured data, or retailers becoming the agents themselves. Watch this space for the reshaping. McKinsey & Company report here: https://shorturl.at/jI6j7 CC Sarah Marzano Collin Colburn Andrew Lipsman Jacqueline Karlin Amelia Van Camp Scott Collins Ryan Verklin Nate Elliott Debra Aho Williamson Invite any hole poking here. Learning all the time. Curiosity in motion. #ai #commerce #retailmedia

Amazon’s “Just Walk Out” technology demonstrates how AI, computer vision, and sensor fusion - similar to what you’d see in autonomous vehicles - can radically transform the way we shop. Instead of waiting in a checkout line, customers simply pick items off the shelf and leave, while the system automatically bills them. Yet with any shift toward automation, there’s the question of how it impacts employment and community economics. If fewer cashiers and clerks are needed, the ripple effects on the local job market can’t be ignored. Additionally, some consumers worry about privacy - advanced tracking systems gather a lot of data on purchasing behaviors, raising concerns over how, where, and for what purpose that data is used. Still, there’s no doubt this is just the beginning of technology-driven changes in retail. Beyond cashier-less shops, we might soon see: - Personalized In-Store Experiences: AI-driven recommendations could pop up on screens or apps as you walk through aisles, guiding you to products based on past purchases or dietary preferences. - Augmented Reality (AR) Fitting Rooms: Try on clothes virtually, see how furniture fits in your living room - without physically moving a thing. - Automated Inventory & Restocking: Shelves that monitor stock levels in real time and reorder items as needed, helping stores avoid both stockouts and over-ordering. - Drone & Robot Deliveries: As last-mile delivery becomes faster, we may see robots handling short-distance delivery or drones zipping packages straight to a customer’s doorstep. What do you think the next big change will be? Is it further automation, more personalization, or something else entirely? #innovation #technology #future #management #startups

Every cart tells a story. A pack of pasta, mozzarella, and fresh basil isn't just three items on a shelf. It's someone planning a date-night dinner. Diapers, wipes, and formula aren't individual purchases. They're an exhausted parent stocking up for the week. At Zepto, one of the fundamental problems our Data Science team obsesses over is: Can we understand what a customer is trying to accomplish before they explicitly tell us? Over the past few months, we built a Cart Contextual Recommendation System that does exactly that. 💡 The Core Shift: From Affinity to Intent Traditional recommendation engines rely heavily on item popularity or pairwise co-occurrences (e.g., "People who bought X also bought Y"). We wanted to move beyond that. We borrowed a powerful concept from Natural Language Processing (NLP): Masked Language Modeling. If large language models learn by predicting missing words in a sentence, why can't recommendation models learn by predicting missing products in a cart? 🛒 A cart becomes a sentence. 📦 Products become words. 🎯 Shopping intent becomes the context. Using a custom Transformer-based architecture trained on millions of historical carts- and enriched with real-time temporal and geographical signals - the model decodes the implicit "mission" of a user: 🍕 Pizza base + Mozzarella → Tomato sauce becomes instantly relevant. 👶 Diapers + Wipes → Baby lotion surfaces as a natural complement. 🥣 Oats + Bananas → Protein powder is predicted as the likely next item. 🚀 The Business & Engineering Impact What makes this incredibly exciting is that the model handles the chaotic, high-velocity nature of quick commerce at Zepto scale. It doesn't just look at what's in the cart; it understands how combinations of products, day of week, and location converge to reveal real-time human intent. The business impact was significant: 📈 Increased Add-to-Cart Rate 💰 Higher Average Order Value (AOV) 📊 Improved Gross Profit Per Order Most importantly, it reduces friction, helping our users complete their shopping missions in seconds. We’ve just published the full engineering deep dive, covering everything from our weighted masking strategies and Transformer choices to our online experimentation framework. Link in the first comment. In modern recommendation systems, predicting what a customer wants is valuable. Predicting why they are shopping is where the real magic happens.

AI impact on shopping is going to be much bigger than ecommerce, because ecommerce is only a small part of retail that starts online. So many buying decisions are driven by interactions online that separating out ecommerce as an isolated channel sales channel is obsolete. To illustrate, Profitero data shows that 64% of retail sales are either ecommerce or digitally-influenced offline retail. This essentially means that even though we still rarely shop on TikTok or YouTube, we buy things because we saw them there. AI tools are now replacing Googling. Our decisions are increasingly digitally influenced by ChatGPT and others. That affects shopping too. And just like the impact of social platforms, AI tools are both direct checkout channels (many have added buy buttons to their responses) and indirect recommendation engines that inform decisions elsewhere. Think someone standing among the endless isles at Walmart chatting with ChatGPT for shopping ideas. AI is disruptive to retail. Ecommerce penetration has been growing slowly in most mature markets and this could re-accelerate it because of better personalization, discovery, and curation. Thus, one way to look at it is what percentage of ecommerce is AI (be it agentic shopping, embeded checkout, etc). Another is what percentage of all retail is influenced by AI. The first is a challenge for Amazon. The second is a challenge for retail.

We just published new consumer AI research, Future Commerce, and discovered something interesting about shopping behaviors that reframes the #agenticcommerce debate. The biggest finding is this: 30 percent of consumers expect to visit physical stores less often because of AI. Among Gen Z, that number climbs even higher. That’s not a small shift. The second signal is deal-seeking. The majority of consumers say they want AI to do the work of finding the best price for them: price comparison, coupon hunting, and retailer hopping. We've been framing the modern consumer as disloyal, but this new data reframes it more as brand fatigue. Consumers are tired of doing the labor themselves. AI is the relief. The third takeaway sits downstream of both. AI is redistributing intent. When an AI tool recommends a product, most shoppers still click through to a brand or retailer site to complete the purchase. Conversion hasn’t disappeared. What’s changed is how prepared to buy consumers are when they arrive. What's it all mean? Put together, these three signals form a clearer picture of the agentic commerce debate. Firstly, AI is not primarily a checkout replacement. It’s an upstream system that reduces friction before the store. It narrows choice and compresses the research done inside of the brand experience, and ultimately shifts influence earlier in the journey. By the time a shopper shows up, the decision is already taking shape. So this is the key Future Commerce insight: cognitive effort is moving away from consumers and into systems, and the first-order impacts will be felt by physical retail. Brand experiences are judged less on inspiration and more on their ability to sustain the momentum of the "collapsed funnel" and the instanteous desire to purchase. Remember: consumers are adapting quickly. Retail infrastructure is adapting slowly. Hit the link for the full research 👉

AI is changing how purchase decisions get made. Consumers are increasingly using tools like ChatGPT, Perplexity, and Copilot to ask “What should I buy?” rather than scrolling through search results. Retailers are already seeing referral traffic from these AI tools, signaling a shift from search‑driven discovery to AI‑driven recommendations. McKinsey projects that AI agents could mediate $3–5 trillion in global consumer commerce by 2030. In this new model, AI agents compare products, summarize tradeoffs, and narrow choices — often before a shopper ever visits a retailer’s site. For retailers, this means discoverability is no longer just about search engine optimization or marketplaces — it’s about being understandable and trustworthy to AI. Winning brands will invest in richer, structured product data, clearer metadata, and machine‑readable content that supports decision‑making. This requires changes across people (AI literacy and ownership), process (continuous product data enrichment), and technology (clean, real‑time, accessible product information). As decisions become automated, the most AI‑discoverable retailers will shape outcomes — everyone else risks being invisible. #RetailTransformation #AgenticCommerce #AIDiscoverable #DataHygiene

I remember presenting to the Bed Bath & Beyond Board in the mid-2000s and walking through how customer behavior was changing. At the time, the shift was driven by search. Customers were no longer starting their journey in a store or with a circular. They were: 🔎 searching on Google 🔎 reading reviews 🔎 asking friends 🔎 and then deciding whether to visit a store or buy online. That model reshaped retail over the next decade. Today, we’re watching another front-door shift.....this time driven by AI and algorithmic discovery. Consumers are increasingly starting with: 👉 “Which air fryer should I buy?” 👉“What’s the best moisturizer for sensitive skin?” 👉“Where can I get this delivered fastest?” And instead of scrolling through pages of results, they’re receiving synthesized answers, ranked recommendations, and direct purchase paths. The implication isn’t just a marketing shift. It’s a structural one. If AI systems and social platforms are now mediating discovery, then: ➡️ product data quality ➡️ content depth ➡️ review signals ➡️ and fulfillment reliability... .....become inputs not just for SEO, but for machine decisioning (now GEO/AEO). I’ve seen this pattern before: when the front door moves, the retailers who recognize it early redesign their operating model around it. The ones who don’t often keep optimizing the last era’s traffic patterns, until they wonder where demand went. #RetailStrategy #AIDrivenCommerce #DigitalTransformation #CustomerJourney #FutureOfRetail