Retailers hit AI shopping trust ceiling

Retail shopping is getting rebuilt around chatbots. But the part retailers really want — the sale — is still stubbornly human. Over the past few months, OpenAI has shifted ChatGPT shopping away from its failed in-chat checkout experiment and toward product discovery, while brands like Walmart, Target, Sephora and Starbucks have rushed to build apps and integrations inside ChatGPT and Claude. The pattern is pretty clear: AI is becoming a recommendation engine, not yet a trusted cashier. (modernretail.co) ### What changed this week? Modern Retail’s snapshot of the market shows retailers and brands launching shopping apps inside ChatGPT and Claude at speed, with roughly 900 ChatGPT apps listed and 353 Claude connectors tracked, and about 10% of ChatGPT apps focused on shopping. But the same piece says adoption and conversion are “pretty (modernretail.co)oogle and into AI assistants. (modernretail.co) ### Why are brands piling in anyway? Because discovery is where the leverage is. If a shopper asks ChatGPT for “best running shoes under $100” or a moisturizer for sensitive skin, the assistant can collapse search, comparison and recommendation into one answer. That means the old fight for homepage traffic matters less, and the new fig(modernretail.co)resence inside the system strategically valuable even before checkout works well. (openai.com) ### What happened to buying inside ChatGPT? Turns out OpenAI already tried the hard version. In September 2025 it launched Instant Checkout with Stripe, Etsy sellers, and promised Shopify merchant expansion. By March 24, 2026, OpenAI had pivoted away from that model after the feature struggled to gain traction. The company said it wanted merchants to use their own checkout experiences instead, while ChatGPT focu(openai.com)ig tell — discovery was easier than transaction orchestration. (openai.com) ### Why is checkout the hard part? Because the moment money moves, trust, liability and control all get sharper. MarTech’s write-up of new consumer data says 77.6% of consumers have used AI to help with shopping in the past six months, and more than 43% do so weekly. But more than half are uncomfortable with AI storing card details, and the most common amount people would let AI spend autonomously is $0. Even fr(openai.com)nce the basket. It still can’t be trusted with the wallet. (martech.org) ### So what are retailers building now? Mostly guided experiences, not fully autonomous ones. Sephora’s ChatGPT app helps with beauty advice and ties into loyalty perks. Starbucks’ beta app lets users ask for drinks by mood or preference. Walmart’s newer ChatGPT integration supports linking accounts, loyalty and Walmart payments, but still keeps the merchant in control(martech.org)eption than the rule. (modernretail.co) ### What does this change for marketing? It creates a second storefront layer — one controlled by AI answers instead of search results pages. Marketers now have to care about whether products are legible to models, whether feeds are current, and whether promotions, reviews and availability can be pulled into AI responses. MarTech’s poi(modernretail.co)on inclusion. (martech.org) ### Is this a temporary trust problem? Maybe partly, but not entirely. People usually adopt AI faster for low-risk tasks than high-risk ones. Asking for gift ideas feels harmless. Letting a bot store your card, choose the merchant, and place the order feels different — more like handing over judgment, not just outsourcing research. That gap may narrow for routine reple(martech.org)ensitive buys. This is an inference from the usage split and the checkout pullback. (martech.org) ### Bottom line? Retailers are not crazy to build for ChatGPT, Claude and similar tools. That is where product discovery is heading. But the current ceiling is trust, not interface design. For now, AI looks like the new front door to shopping — not the register.