Source Policy

1. Where the data comes from

Clinic records and reviews are gathered from publicly accessible platforms. We do not collect member-only material or anything behind a login.

• Google Maps: review bodies, ratings, post dates, author display names, and the platform review total
• Naver: listing data. There is no review-text API, so counts are maintained by our operators
• Clinic websites: factual fields such as address, opening hours, and medical staff
• Gangnam Unni, Babitalk, Kakao Map: no official API exists, so we publish only what an operator has verified individually

2. How the review analysis is produced

Collected review bodies are passed to a language model in batches of 50 to extract recurring themes. Results are consolidated per clinic and shown as strengths, cautions, per-aspect scores, and sentiment distribution.

Every summary is generated from real review text. We never write a summary for a clinic that has no reviews.

3. What we exclude

None of the following is used as evidence.

• Promotional language: advertising copy, discount campaigns, before/after marketing claims
• Posts too short to carry any judgement
• Posts whose content is unrelated to the clinic in question
• Paid placement or partnership. We take no listing fees from clinics

4. Reading the counts

"Analysed reviews" is the number of review bodies we collected and used. "Platform review total" is what the source service publishes. The two do not match; the former is always smaller.

5. Confidence tiers

Each analysis carries a confidence tier. Low counts, a single source, or stale newest-review dates all push confidence down, and the result is shown as an early signal that needs verification. Treat low-confidence entries as hypotheses to check, not conclusions.

6. Limitations

Public reviews are structurally biased. People at both extremes of satisfaction post more readily, and foreign patients post less than local ones. Reviews also cannot speak to whether a procedure was medically indicated or technically sound.

Our analysis is neither a quality guarantee nor a recommendation. It is material for deciding what to ask during a consultation.

7. Corrections and removals

Factual errors, outdated records, and delisting requests go to kanbikorea@gmail.com. When a clinic contacts us, we respond citing the source the entry was based on.