We answer with measurements, not vibes. We test whether AI crawlers can read your site and whether AI answers mention you — then prescribe, in priority order, what will change that.
| K-beauty export brand | Score | Critical issue |
|---|---|---|
| Brand M (US entity) | 53.0 | Best of cohort — still short |
| Brand B | 49.5 | Zero citable paragraphs |
| Brand A | 30.0 | Blocks GPTBot (403) |
| Brand R | 21.1 | Blocks GPTBot (403) |
| Brand ML | 3.0 | Blank page to AI (SPA) |
| GEO-optimized baseline | 85+ | — |
On July 9, 2026 we audited ten export-focused K-beauty brand sites with one reproducible method: fetch each page as a normal browser, then as OpenAI's crawler, and compare. Three sites returned 403 to GPTBot while serving browsers normally — almost certainly CDN security rules misfiring. Two more render entirely client-side, so AI crawlers see an empty page. Not one site in the cohort had a single paragraph shaped the way AI answers actually lift text (a self-contained answer of roughly 134–167 words). The category is wide open: the first brands to fix retrieval and publish answer-shaped content will take the recommendation slots for everyone's buyer queries.