Writing about faces is easy to get wrong. This document spells out the language rules we use across result cards, blog posts, and marketing copy. The goal is to keep FaceOracle useful and entertaining without drifting into claims it can't support.
Language we use
- Mood, vibe, impression, style. We describe the visual feel of the image itself.
- For fun, entertainment, creative. These words flag that a result is not a scientific judgment.
- "Traditionally interpreted as…" for cultural material such as physiognomy.
- "Studies have reported a tendency toward…" when we cite first-impression psychology — tendencies, not rules.
Language we refuse
- Deterministic personality claims — "You are a(n) ___ person."
- Sensitive-attribute inference — "The AI can tell your gender / nationality / race / health."
- Predictions that invite discrimination — criminality, hiring outcomes, creditworthiness.
- Ranking phrases like "the ideal face" or "faces to avoid."
- Scientific certainty — "100% accurate," "proven by science" — because we cannot back those claims.
Result-card rules
- Every result page shows an entertainment banner at the top and the bottom.
- Each card states what it does not determine — personality, ability, health, identity.
- Similar-vibe celebrity cards carry the line: "Not face recognition or identity verification, not a claim of actual resemblance."
- Terms of Service prohibit using results for hiring, admissions, lending, contracts, or security access.
Blog rules
- Every post shows category and reading time; entertainment-only pieces are labeled as such.
- We use public-source imagery (Wikipedia, Naver Search) when we reference specific people.
- Creative content (past-life jobs, animal look-alikes, couple mood stories) is clearly framed as fiction inside the article so readers never mistake it for analysis.
Why we bother
"AI reads a face" sounds authoritative, but what an image model really does is describe the visual features of a photo in words. If we hide that difference, readers may end up judging themselves or others by a standard that simply doesn't exist. Softening our language is a small cost to pay for keeping the service honest.