The first cross-platform benchmark for AI commerce. We scanned 2,400 commercial queries across ChatGPT, Gemini, Claude, Perplexity, and Microsoft Copilot — and verified every recommended product against its canonical U.S.-facing PDP. This is what we found.
Buyability scores across the 12 most-recommended Korean beauty brands in U.S. AI shopping. Per-platform breakdowns highlight where each brand wins or loses execution.
Scoring methodology: queries weighted by U.S. AI shopping volume (Bain, March 2026 benchmark). Per-platform scores reflect successful checkout-path validation against the canonical U.S.-facing PDP. See full methodology.
Across the bottom-half cohort, broken-impression rates concentrate in two failure modes: inventory drift and price disagreement across multi-listing footprints.
Most common in seasonal SKUs and limited Olive Young exclusives that AI agents continue to recommend after sell-through.
Share of failures · 31%Same SKU sold at four different prices across YesStyle, Soko Glam, Sephora.com, and Amazon — agents select inconsistent canonicals.
Share of failures · 27%Pages that fail Open Graph or schema.org validation are silently down-ranked by agent retrievers — invisible to most brand teams.
Share of failures · 17%Hangul-only SKUs surface in U.S. results because agents see Korean catalog data without market-availability gating.
Share of failures · 15%Checkout cannot complete to a U.S. address; the link routes to a grey-market reseller.
Share of failures · 10%Buyability varies meaningfully by AI surface. Brands optimizing for a single platform are leaving the equivalent of one full surface's worth of conversion on the table.
Run a free Scanner pass to see where you appear in AI today, and how your buyability score compares to category leaders. Onboarded in 48 hours.