Ask a customer to describe the scent they want and you will often hear: “I don't know the notes, but I love the smell of that one perfume.” Traditional online stores cannot do much with that. Keyword search only finds products whose titles contain the words you typed. Fragrance needs something smarter.
Similarity, not keywords
AI recommendation turns each fragrance into a profile: its main accords, its top, heart and base notes, how sweet or fresh it is, and how it is described by people who wear it. Once every perfume and every oil has a profile, finding a match becomes a question of distance: which oils sit closest to the perfume you named?
Three ways to ask
- By reference: “smells like a famous blue men's fragrance”. The system looks up the reference profile and ranks the nearest oils.
- By note or mood: “rose and oud for weddings” or “fresh aquatic for summer”. It filters and ranks by accords and occasion.
- By budget: combine either of the above with a price ceiling per kilo.
Why it beats browsing
In a library of thousands of oils, browsing by category quickly becomes exhausting. Similarity search compresses hours of sampling into seconds, and it surfaces options a buyer would never have found by title alone, such as an original composition that happens to sit very close to a well-known perfume.
Its limits, honestly
An AI match is a strong shortlist, not a guarantee. Two perfumes can share a note pyramid and still smell different because of material quality and proportions. That is why we describe our inspired oils as bench-matched to 85–90% and always recommend sampling before a large order.
Getting better results
- Name the exact perfume and, if possible, the edition (eau de toilette and parfum versions can differ).
- Add what you don't want: “oud without leather”.
- Tell it the use: fine fragrance, attar, candle or soap.
Ask Haveli AI now, or tap the gold chat button on any page. You can also photograph a bottle with our Perfume Decoder.