The 2026 thesis
AI engines research products before shoppers do.
The product-research session has moved. Shoppers ask ChatGPT and Perplexity which product to buy, Google folds AI Overviews and AI Mode into the results page, and pre-purchase agents — ChatGPT Shopping, Perplexity's Buy with Pro — draft carts on the buyer's behalf. Checkout protocols such as the Agentic Commerce Protocol (OpenAI/Stripe) and Google's Universal Commerce Protocol are in pilot as of mid-2026. A brand can lose the sale before its site logs a visit.
Most tooling answers this with brand-level mention counts. But the answer engines don't recommend brands; they recommend products. A brand mention can't say which SKU won, which PDP an agent failed to parse, or what to change on Tuesday. eCommerce Insights watches the unit of revenue instead: every product tracked across ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, and Copilot, each PDP scored twice — citation and agent-readability — and every failing page given its title, schema, and metafield fixes as a reviewable diff a human approves.
The product-level obsession is inherited. Indellia's first platform connects consumer-brand feedback — reviews, support tickets, returns, surveys — to the specific SKU it describes, because feedback detached from the product it's about is noise. eCommerce Insights applies the same discipline to AI visibility, where the stakes are now pre-purchase: the engines and agents that research, shortlist, and increasingly carry products to checkout. The category context is in the product AI visibility guide; the mechanics live in product-level (SKU) tracking.
