Stanford AI Index: 67% of US shoppers have used an AI assistant to research a purchase

Stanford's Human-centered AI Institute published its annual AI Index. The shopping-behaviour chapter reports that 67% of US online shoppers have used an AI assistant — ChatGPT, Perplexity, Claude, or Gemini — at least once to research a purchase in the last 12 months. The number is 54% in the EU, 71% in India, and rising in every market surveyed.

Apparel, consumer electronics, and home goods categories show the fastest growth in assistant-mediated research. Consumers report that the assistant's synthesised answer is often the final source of truth, even when they click through to a merchant site to complete the purchase.

When two-thirds of US shoppers research via an AI assistant, the citation set of those assistants becomes the shelf. If your Shopify store isn't in the citation set for the 'best X for Y' prompts your buyers type, you're not on the shelf — even if your SEO rankings are great.

The research also confirms the asymmetry the Surfient thesis has assumed: the assistant sends pre-qualified, high-intent traffic to merchants it cites. A shopper who arrives via a Perplexity or ChatGPT answer has already read the synthesised recommendation. Their click is downstream of a decision, not the start of comparison.

The practical takeaway is to audit which prompts your buyers ask (Surfient's Prompt Intelligence module does this), and confirm your store is in the citation set for the top 10 to 20 of them. Where it isn't, fix the specific retrieval gaps — missing schema, missing answer blocks, missing llms.txt entries.