SkinSeoul app in ChatGPT: privacy notice

Last updated: November 9, 2026

This notice explains what happens to data when you use the SkinSeoul app (plugin) in ChatGPT. It supplements the SkinSeoul Privacy Notice. The app is operated by SkinSeoul Pte. Ltd. It lets ChatGPT search SkinSeoul's public product catalog. It does not have access to your SkinSeoul account, orders, payment details, or your ChatGPT conversation, and it does not place orders.

What we receive

When ChatGPT calls the app, we receive only the following:

DataExampleWhy we receive it
Your request to the toolsearch words (up to 200 characters), category, brand, skin type, skin concern and ingredient filters, price range, product IDs, currency, page size, recommendation typeTo find matching products
ChatGPT's anonymized identifiers (openai/subject, openai/session)a random-looking ID created by OpenAI; it is not your name, email or OpenAI account IDTo apply fair-use rate limits
Network informationIP address of the server that sent the request (for ChatGPT, an OpenAI server), user agent, time, request path, response statusTo operate the service, keep it secure and troubleshoot errors

Skin type and skin concern filters describe the products you are looking for. We do not use them to identify you, and we do not keep them in any profile.

The app does not ask for your name, email address, phone number, postal address, payment information, account credentials, precise location or the rest of your ChatGPT conversation. If you type any of these into a search, they are handled like any other search words: they reach our catalog service and can appear in its request logs and diagnostics for the periods below. ChatGPT may also send a locale, an approximate location, a user-agent hint and an anonymized organization ID; the app ignores them and does not store or log them.

How we use the anonymized identifiers

We accept these identifiers only from OpenAI's published connector IP ranges. As soon as a request arrives, we replace each identifier with a keyed one-way hash (HMAC-SHA-256). We never store or log the original identifier. The hash is kept only in the app's server memory. We use it only to count requests for rate limiting (one-minute windows). We do not use it to create a profile or a visitor ID, and we do not send it to any other service. It is never written to disk or logs, and it is discarded when the server restarts or when newer entries need the space.

What the app returns to ChatGPT

The app returns public catalog information only: product ID, name, brand, categories, description, size, regular and sale price, currency, stock status, product page link, up to three product image links, and the time the data was retrieved. Product details can add the product's skin-type, skin-concern, benefit and key-ingredient tags and usage instructions. Routines add step numbers and names. Skin-filter lists return filter IDs, names and slugs.

The app never returns your identifiers, any data about you, internal IDs, or diagnostic or debugging data. If something goes wrong, it returns only an error type and a fixed message. Every field the app can return is fixed in our code, including nested fields, and automated tests check this.

Product links include utm_source=chatgpt&utm_medium=ai_agent so we can see that a visit came from ChatGPT. If you open a link, the SkinSeoul Privacy Notice (including cookies and analytics) applies to your visit to skin-seoul.com.

Who receives the data

We do not sell this data, use it for advertising, use it to train AI models, or combine it with SkinSeoul customer accounts.

How long we keep it

DataWhereRetention
Hashed identifiers for rate limitingapp server memory onlyuntil the server restarts or the entry is replaced (memory holds at most 10,000 entries); never written to disk
Network information (load balancer access logs)Google Cloud Logging30 days
Your tool request (search words, filters, product IDs, currency) in our catalog service's request logsGoogle Cloud Logging30 days
Error reports and sampled performance traces from our catalog serviceSentryerror reports up to 90 days; sampled performance data up to 13 months, depending on our Sentry plan

Your choices