How It Works
What happens when you search.
When you submit a description, it goes directly to an AI model — the same kind that powers large language assistants. Attached to that request is the full library: every fragrance we have, with its notes, description, character, occasions, and a set of descriptive keywords. The AI reads all of it, ranks every fragrance against your words, and returns three: the best match and two alternatives. Nothing is ranked by popularity or price. The AI reads the same set of fragrances every time, in the same way, and the result reflects only how well each one fits what you described.
How accurate is it, honestly.
Better than random. Worse than a nose. The AI has never smelled anything. What it knows about fragrances comes from structured data — notes, descriptions, and language used by perfumers and critics. It's good at matching mood, occasion, and character. It struggles when the distinction between two fragrances is subtle and sensory: a specific drydown, a texture, the way a note shifts on skin over hours.
The match score is a confidence estimate, not a guarantee of liking. A 91% match means the AI is highly confident the fragrance fits your description. It does not mean you'll love it — fragrance is too personal for that. Treat the result as a starting point: a name worth seeking out at a counter, not a verdict to buy blind.
The library — what's in it and why it matters.
Every recommendation comes from what's in the library. If a fragrance isn't in it, we can't recommend it — no matter how perfect the match might be. Each entry includes notes, concentration, price, seasons, occasions, sillage, longevity, and a set of keywords: descriptive phrases that map natural language to scent character. "Boardroom confident." "Rainy Sunday." "Something that doesn't announce itself." The keywords are what the AI primarily reads. A fragrance with thin keywords will rank lower than it deserves. One with rich keywords will surface when it should. We add new fragrances and improve keyword coverage regularly.
One recommendation. Clear reasoning.
Olfome returns one primary match — the one it's most confident about — and two alternatives for comparison. Each comes with a match score, a plain-language explanation of why it fits your description, the scent's notes, its price, and a direct link to buy it.
Refine until it's right.
If the first result isn't quite what you had in mind, tell it. "More subtle." "Under $80." "Something less well-known." The conversation continues until you've found your scent.
This is free. Here's how, and why it stays honest.
Olfome is free to use. We don't charge you, and no brand has ever paid us to appear in results.
We make money through referral links. When you click "Buy" and make a purchase, we may receive a small commission from the retailer. That commission is attached after the AI has already made its decision. The perfume at the top is there because an algorithm chose it — not because a brand paid for the placement. We don't touch the ranking. We don't review it. We don't adjust it. The AI runs, the result comes back, and we attach a link to wherever you can buy it.
The AI makes mistakes. Fragrance is personal, and a match score is not a guarantee. But the mistake is made by an algorithm with no financial relationship to any brand — not by a salesperson on commission, a magazine funded by advertising, or a bestseller list shaped by shelf deals. We think an honest mistake beats a biased one. That's the whole idea.