kbeautytmi › AI citations

When people ask AI about K-beauty,
whose name comes back?

Every week we ask 4 AI engines (Perplexity · Gemini · OpenAI · Claude) the same 10 K-beauty questions — best toner for sensitive skin, best Korean sunscreen, and so on — and count which brands the answers actually cite. Citations ≠ sales. This is the discovery layer forming before anyone reaches a shelf. Measured 2026-07-27

The leaderboard

Violet bar = how often AI names the brand. The chip = how many of our 20 retail channels it actually charts on — the gap between fame and shelf is the story.

1cosrx 124on 12 channels
2anua 92on 15 channels
3dr.jart 72on 5 channels
4etude 68on 5 channels
5purito 42on 7 channels
6skin1004 40on 10 channels
7torriden 38on 5 channels
8round lab 31on 12 channels
9pyunkang yul 21on 2 channels
10beauty of joseon 19on 10 channels
11isntree 10on 3 channels

Week over week

2026-07-06
№1 COSRX 40
2026-07-13
№1 Anua 36

The trend line is the product — the archive deepens every Monday and Thursday.

Why this matters

Shoppers increasingly ask AI before they search, and AI answers concentrate on a handful of names. A brand that dominates retail shelves but never gets cited is invisible in the next discovery channel — and a brand AI loves that hasn't hit mass retail is a signal buyers can act on. Cross-reading this leaderboard against our retail ladder is the whole point: in 2026-07, the top of the AI leaderboard and the top of the retail-reach ladder were the same brands — evidence the discovery layer and the shelf are moving together.

Method: identical prompts to 4 engines, brand mentions extracted from answers with token-boundary matching against our chart-brand dictionary. Reprocessed brand-neutrally by kbeautytmi. AI citation ≠ endorsement, quality, or sales.