Vetted Targets Messy Product Research
Stuart Kearney, co-founder of Vetted, on AI agents in shopping
The company is betting that whoever cleans up product research wins the whole shopping relationship. Research is where shoppers hit too many options, too many sponsored placements, and too little trustworthy synthesis, so the winning product is not just a search box, it is a guide that asks follow up questions, narrows the field, compares tradeoffs, and earns enough trust to later own discovery above and checkout below.
-
This is why going upstream is harder than starting at checkout. Checkout is mostly payments, merchant integrations, fraud, and order confirmation. Research adds taste, intent, budget, use case, and explanation, which is why Vetted uses multi turn conversations and a large orchestration layer of model calls to shape decisions before the click.
-
The clearest comparable is Honey. Honey worked when the user already knew what to buy and just wanted a coupon or faster purchase path, but it struggled to become the place people began their shopping. Vetted is trying to solve the earlier moment, when the shopper still needs help deciding between the 10 pack, 30 pack, wood, plastic, cheap, premium, and fast shipping versions of the same basic item.
-
That position matters because the research layer is where ad heavy marketplaces and broad AI assistants are weakest. Amazon is reliable at fulfillment, but crowded with paid placement. OpenAI and Perplexity can show products, but Vetted argues that one click buying still breaks when merchant data, stock, pricing, returns, or SKU level details are wrong. The control point is not payment, it is trusted narrowing.
The next step is a blended shopping interface where AI handles the messy decision work first, then hands off to either a merchant or a native checkout flow once intent is precise. If that transition gets smoother, research products become the new front door to commerce, and marketplaces, payments companies, and LLM apps will all race to own that layer.