How Our Recommendations Work
Our recommendations come from a three-part system: your answers, our catalog, and an AI matching layer. Here's exactly what happens when you take the quiz - including what the AI can and cannot do.
Step 1 - Your preferences
The quiz asks four things: who the fragrance is for, the scent styles you're drawn to (shown as images, not jargon), the occasions you have in mind, and whether you prefer designer originals or inspired-by alternatives. We do not ask how strong or long-lasting you want it: we have no sourced strength or longevity data yet, so we could not match on it honestly.
Step 2 - Catalog filtering
Before any AI is involved, our system filters and ranks our quiz catalog - more than 200 fragrances - against your answers. Notes are being checked against brand and retailer sources page by page; each perfume page shows whether its notes have been checked, when, and against which sources. This scoring is deterministic: shared scent families and matching notes earn points. Occasions are passed to the AI as context so it can consider your intended use, but they do not influence the numeric score - they are editorial estimates, not verified attributes.
Step 3 - AI matching, with guardrails
An AI model then reviews only the top-ranked candidates from our catalog and selects the handful that best fit your overall profile, explaining each choice in plain language. The important part: the AI can only choose from the catalog we give it. It cannot invent a perfume, hallucinate a price, or recommend something that is not in our catalog. Every recommendation you see is cross-checked against our database before it reaches your screen.
What we don't do
We don't accept payment for placement in recommendations. We don't alter results to favor any brand. And we don't pretend the AI is a perfumer - it's a matching tool working from real data, reviewed by people who care about fragrance.