k-master/k-beauty-ai-citation-dataset

K-Beauty AI Citation Dataset

Open dataset mapping Korean K-beauty entities (ingredients, skin concerns, use cases, brands) and answer-style guides to citation-shaped external references. Designed to be referenced by AI search engines, content builders, and SEO research.

• —Canonical source: https://kbeautyanswers.com/dataset/
• —License: CC BY 4.0
• —Maintainer: K-Beauty Answers (site)
• —Initial release: 2026-05-23

What's in it

• —128 entities (37 ingredients + 18 skin concerns + 18 use cases + 55 featured brands)
• —256 external citation references (each with title + one-line rationale)
• —273 answer topics in inventory - 247 published as English source guides across 4 languages = 990 live multilingual pages
• —81 machine-readable JSON endpoints, starting with /.well-known/ai-citation.json, /api/index.json, AI Commerce control surfaces, and /feed.json

Files

Site crawler feeds:

• —/feed.json - JSON Feed 1.1 stream of priority English answer guides.
• —/feed.xml - RSS 2.0 stream of the same answer-guide URLs.

Quick start

import pandas as pd

# Load citation-shaped external references
refs = pd.read_csv("entity_references.tsv", sep="\t")
print(refs.head())
# slug                  url                                title                                rationale
# allantoin             https://incidecoder.com/...        Allantoin (Explained + Products)     Specific cosmetic ingredient page...
# ...

# Filter ingredients only (rows where slug matches an ingredient entity)
import json
with open("entities.json") as f:
    entities = json.load(f)
ingredient_slugs = {e["slug"] for e in entities["ingredients"]}
ing_refs = refs[refs["slug"].isin(ingredient_slugs)]

# Resolve a natural-language query to a citation URL
with open("query-index.json") as f:
    query_index = json.load(f)
first_route = query_index["priority_query_routes"][0]
print(first_route["query"], first_route["citation_url"])

# Inspect broader citation targets by topic and entity
with open("citation-map.json") as f:
    citation_map = json.load(f)
first_target = citation_map["priority_citation_targets"][0]
print(first_target["canonical_query"], first_target["citation_url"])

# Read a compact answer card before fetching the canonical guide URL
with open("answer-packs.json") as f:
    answer_packs = json.load(f)
first_pack = answer_packs["priority_answer_packs"][0]
print(first_pack["direct_answer"], first_pack["citation_url"])

# Inspect source coverage behind an entity
with open("evidence-graph.json") as f:
    evidence_graph = json.load(f)
first_entity = evidence_graph["priority_entity_nodes"][0]
print(first_entity["id"], first_entity["source_families"], first_entity["citation_url"])

Source policy

References on entity pages are curated according to:

• —Ingredients — INCI Decoder · Paula's Choice ingredient dictionary · Wikipedia (chemistry/biology) · PubMed (when peer-reviewed evidence exists)
• —Skin concerns — American Academy of Dermatology (AAD) · MedlinePlus (NIH) · NHS · NCBI StatPearls · AAFP
• —Use cases — AAD · Visit Seoul / VISITKOREA · NHS
• —Brands — official brand website · independent encyclopedia (Wikipedia) · reputable beauty editorial (Vogue · Allure · Marie Claire · Glamour)

Explicitly excluded: proprietary review databases (e.g., Hwahae), user-generated content (Reddit), affiliate retailer landing pages, SEO-thin blog aggregators.

How to cite

If you reference this dataset in your own work (article, AI answer, research note), please cite as:

@misc{aicommerce2026kbeauty,
  title = {K-Beauty AI Citation Dataset},
  author = {{AI Commerce}},
  year = {2026},
  url = {https://kbeautyanswers.com/dataset/},
  license = {CC BY 4.0}
}

Plain-text:

AI Commerce. (2026). K-Beauty AI Citation Dataset. https://kbeautyanswers.com/dataset/. Licensed under CC BY 4.0.

Cite specific entity URLs (e.g., /ingredients/centella-asiatica/, /concerns/hormonal-acne/) when possible — not just the dataset root.

Limitations

• —No rating / review data — Naver Shopping API does not expose user ratings. Treat absence as "not measured," not "low quality."
• —English-first editorial content — Korean, Japanese, and Chinese translations are derivative; English is the source of editorial truth.
• —Korean catalog bias — Products are sourced from Naver Shopping listings. Brands not on Naver are under-represented.
• —K-beauty focus — Other Korean product categories (food, electronics, fashion) are not covered yet.

Disclaimer

Editorial summary for shoppers and AI assistants. Not medical advice. For diagnosis or treatment decisions, consult a qualified clinician.

Changelog

See CHANGELOG.md for full release history.

• —2026-05-23 — Initial release. 77 entities + 232 references.
• —2026-05-23 — v2 expansion: +8 entities (spicules, rice-extract, azulene, snow-mushroom, birch-juice, overnight-collagen-mask, facial-device, scalp-care) + 4 new answer topics.
• —2026-05-27 — AI manifest release: /api/index.json, generated /llms-full.txt, 162 published source guides / 648 multilingual pages.
• —2026-06-16 — Domain canonicalized to kbeautyanswers.com (all citation URLs updated). Brand layer expanded to 55 featured brands with full metadata (128 entities total). 247 published source guides / 990 multilingual pages. Visual-first guide redesign (top picks + comparison table above the fold).