Yes and no. Artificial intelligence can now suggest perfume formulas and predict how a molecule is likely to smell, and perfumes developed with its help have been sold. But in every documented case, people chose what to make, smelled the results, adjusted the formulas and decided what reached the bottle. AI is used as a tool by perfumers and fragrance companies; it does not smell anything itself, and "made by AI" on a label usually describes a collaboration rather than a machine working alone.
What AI has actually done in perfumery
Suggesting formulas
In 2018 the fragrance company Symrise and IBM Research announced Philyra, a system trained on formulas and data that proposes new combinations of materials for a given brief. Its first commercial use was two perfumes for the Brazilian beauty company O Boticário, launched in 2019. Reports at the time described a Symrise perfumer reviewing and adjusting Philyra’s proposals — for example to emphasise certain notes and improve diffusion — before the perfumes were finalised. Symrise later presented an updated version of the system.
Other large fragrance houses have described similar tools for their perfumers, which suggest formulas, check them against constraints such as cost and regulation, or help explore variations.
Predicting how molecules smell
In 2023 researchers published a study in Science describing a "principal odor map": a machine-learning model, trained on thousands of molecules with odour descriptions, that predicts what a molecule will smell like from its structure. On the test set used in the study, its predictions were comparable to those of trained human panellists for many molecules. Several of the authors were from the company Osmo, which grew out of research at Google. Tools like this may help chemists search for new aroma molecules faster.
What AI does not do
- It does not smell. Every prediction is learned from human descriptions of smells. Whether a perfume is beautiful is still judged by noses.
- It does not replace safety assessment. Every perfume sold must meet safety standards and the law of each market, whoever or whatever suggested the formula. See How Is Perfume Made?.
- It does not make a perfume original on its own. A model trained on existing perfumes tends to suggest variations on what already exists; originality still depends on the brief and on human choices.
AI is a new kind of tool in the perfumer’s hands — closer to a very fast assistant than to a perfumer.
How to read "AI perfume" claims
When a brand says a perfume was "created by AI", it is worth asking:
- What did the AI do? Suggest ingredients, generate a starting formula, or help choose from consumer data?
- Who smelled and finished it? In documented cases, a perfumer did.
- Is it just marketing? "AI" can describe anything from a serious formulation tool to an online quiz that recommends existing perfumes.
None of this makes such perfumes better or worse. As always, the only test that matters is how a fragrance smells on your skin. See How to Test a Perfume Before You Buy It.
Will AI replace perfumers?
Nobody can say for certain, and predictions should be read sceptically. What the record shows so far is that AI has changed some of the work — searching, suggesting, checking — while perfumers continue to create, smell and decide. The craft has absorbed new tools before: synthetic molecules transformed perfumery in the nineteenth and twentieth centuries without ending it. See How Synthetic Molecules Changed Perfume and What Does a Perfumer Do?.
Common questions
Has a perfume been made by AI?
Perfumes developed with AI tools have been sold, such as the two made with Symrise’s Philyra for O Boticário in 2019, but perfumers reviewed and adjusted the formulas.
Can AI predict what a molecule smells like?
Research published in 2023 showed a model that predicts odour descriptions from molecular structure with accuracy comparable to trained panellists on many test molecules. It learned from human descriptions; it does not smell.
Are AI perfumes safe?
They must meet the same safety rules as any other perfume. How a formula was suggested does not change the requirements.
Sources
Symrise and IBM Research, announcements of the Philyra system (October 2018); reporting on the O Boticário launch (2019). Lee, B. K. et al., "A principal odor map unifies diverse tasks in olfactory perception", Science 381 (2023), 999–1006.
Further reading
For perfumers, the RUŌOD Lab Academy’s lesson on using AI in research and formulation sets out its uses and limits at the bench. RUŌOD Lab’s own formulation app contains no AI: Perfume Formulation Software Guide explains what formulation software does.