Findable파인더블
Start free
Findable Industry Report · 2026 Q2Five K-beauty brands · Seven AI engines

K-Beauty GEO Report
Where did brands appear in Q2 2026?

We measured Medicube, Round Lab, Anua, Beauty of Joseon, and d'Alba across ChatGPT, Claude, Perplexity, Gemini, HyperCLOVA X, Naver, and Daum. We report those observations separately from editorial techniques tested in Princeton's KDD '24 GEO study.

Methodology note: All five are public K-beauty D2C brands, not Findable customers. Our measurements analyze public AI answers. The reported simulation figures come from Princeton's academic experiments; they are not observed outcomes for these brands.
1.0 · Headline Metrics

Five brands at a glance.

93.2/100Average K-beauty SoVFive brands measured
100%Mention rate in four global enginesChatGPT · Claude · Perplexity · Gemini
55%Average Daum mention rateAcross the five brands
+41%Relative lift from quotationsKDD '24 Table 1 experimental mean
2.0 · Key Insights

Four patterns in the five-brand sample.

01

K-beauty brands appeared consistently in global AI answers

The five brands averaged 93.2/100 in Share of Voice (SoV), with Anua highest at 96/100. Brand mentions appeared in responses from all four global engines measured. This dataset does not establish whether distribution or reviews caused those mentions.

Action: Check whether official English product information and verifiable third-party sources agree on the facts. Compare cited domains in the next measurement.
02

All five brands appeared in HyperCLOVA X responses

Each of the five brands was mentioned by HyperCLOVA X for the four questions measured. This result is limited to those dates and questions; it does not stand in for Naver Search or AI Briefing visibility.

Action: Measure Naver Search and HyperCLOVA X separately. Prioritize content according to the sources each answer actually uses.
03

Daum mentions were lower, but varied by brand

Daum's average brand mention rate was 55% across the five brands. Anua reached 75%, so it would be inaccurate to say every brand scored 50% or lower. This measurement alone cannot identify the cause.

Action: Compare the questions with and without a Daum mention and examine their cited sources. Fill gaps in the official pages, then measure again under the same conditions.
04

Search results and AI answers need separate metrics

Search impressions and brand mentions in AI answers come from different measurement methods. This dataset did not measure revenue, so a visibility gap should not be presented as a sales gap.

Action: Save a baseline using the same question, engine, and region. Make one editorial change at a time and record the difference after remeasurement.
3.0 · Apply 5 StrategiesPublished evidence + questions to test

Editorial methods tested in research—and what remains to test.

⚠ Scope of evidence. The +27%, +41%, and +31% figures are mean relative improvements in the primary metric reported by Princeton's English-language KDD '24 GEO-Bench experiments. They are not results for these five brands and do not predict outcomes in Korean. The other two items are questions for future validation, not numerical forecasts.
S1+27%

Cite Sources

Mean relative improvement for citing original sources in the paper's experiments.

S2+41%

Quotation Inclusion

Mean relative improvement for including quotations in the paper's experiments.

S3+31%

Statistics & Data

Mean relative improvement for adding statistics in the paper's experiments.

S4To validate

Korean Entity Grounding

Unify Korean, English, and mixed-script brand names, then measure before and after.

S5To validate

AI Briefing Visibility

Measure Naver AI Briefing appearances in a separate sample.

4.0 · Related Assets

Related resources.

K-GEO-Bench v0.1 (open dataset)

Raw data for 140 measured responses. Download JSONL or JSON under CC BY 4.0.

Case study: Before/after simulation

A clearly labeled simulation of possible GEO changes for five brands.

Start in 3 minutes

How does AI answer about
your brand right now?

Just your email, done in 3 minutes. No card required.

Start freeBook a call