Artificial Intelligence and The Science of Predictive Perfuming
For centuries, perfumery has existed somewhere between chemistry and poetry. A perfumer works with molecules, certainly, but also with memories, emotions, cultural associations, and the mysterious relationship between fragrance and human skin.
Now, artificial intelligence is attempting to turn this most elusive of senses into something measurable, predictable, and programmable.
Companies developing AI-generated perfumes claim that machine learning can dramatically accelerate fragrance creation, identify new scent molecules, and even translate cultural trends into wearable fragrances.
At the forefront of this emerging technology is Osmo, a company developing what it calls Olfactory Intelligence. Its foundational technology, the Principal Odor Map, uses artificial intelligence to predict how molecules will smell based on their chemical structures.
But while AI may become extraordinarily good at predicting the odor of a molecule, it faces a much greater challenge: understanding the deeply individual experience of wearing perfume.
Can an algorithm predict how a fragrance will interact with your skin chemistry? Can it account for genetic differences in our sense of smell? Can it understand why a particular scent awakens a forgotten memory, or why something one person finds beautiful may be unpleasant to another?
And perhaps most importantly, what happens when the fragrance industry begins using artificial intelligence to turn our cultural anxieties, emotional desires, and social media behaviors into products designed for us before we even know we want them?
What Is the Geometry of Smell? Understanding AI's Principal Odor Map
Traditionally, scientists have struggled to predict exactly how a molecule will smell simply by examining its chemical structure. Two structurally similar molecules can produce remarkably different odors, while molecules with very different structures may smell surprisingly alike. Using machine learning and molecular data, AI moves from organizing fragrance ingredients according to their chemical structures and instead uses a model of relationships (The Geometry of Smell) between molecular structure and human odor perception. Molecules predicted to smell similar are positioned near one another within a multidimensional mathematical map.
In a landmark 2023 study published in Science, researchers evaluated the model against a trained human sensory panel. Across 400 previously uncharacterized odorants, the model's predictions more closely matched the average panel assessment than those of the median individual panelist. While this was a significant scientific achievement, the study also revealed a fundamental limitation.
The AI was predicting the average response of human noses, not the experience of any particular human being. And in perfumery, that distinction is everything.
How AI-Generated Perfumes Turn Social Media Trends into Fragrances
Companies are now exploring ways to use AI to transform images, written descriptions, brand identities, and cultural trends into proposed perfume formulas. Platforms like Osmo's Olfactory Intelligence platform allow creative concepts to be translated into fragrance compositions without utilizing human perfumery expertise. In 2025, this translated the social-media-driven Fisherman Aesthetic into a fragrance called Where Water Ends - a mood board featuring rocky coastlines, cable-knit sweaters, tidal pools, and maritime imagery became a composition of marine notes, aldehydes, soft florals, and woods. In 2026, trend forecasting company WGSN identified a cultural response to widespread stress, uncertainty, and overstimulation and a craving for nostalgia, playfulness, and escapism and named it the Unserious Everything macro trend. AI translated this into a fragrance featuring cotton candy, berries, maple syrup, balsamic accords, and musks.*
* In comparison, Alkemia's take on this collective expression of human emotion was the perfume Lacrimae Rerum ("The Tears of Things") - a poignancy of salt and skin, crushed nicotania flowers, purple irises, Oregon Grape leaves, bitter orange, silver needle tea, guiacwood, weeping cypress, Isatis bark, and a sweet immolation of cloves.
The implications of AI on perfuming is significant.
Imagine a world in which a particular aesthetic begins trending on social media on Monday. By Tuesday, a company or brand has identified its visual language, emotional associations, and commercial potential. AI has generated a proposed fragrance formula designed to embody those associations. By Wednesday, perfumers using AI refine the composition, and manufacturers prepare a product for an audience already primed to desire it. By Thursday, AI has generated a comprehensive marketing plan to leverage maximum profit from the collective pain of humans feeling disconnected, stressed, unable to cope, and increasingly vulnerable.
Perfume is being repositioned from the art of desire into a calculated technology that follows, predicts, and monetizes.
1. There is No Universal Human Nose: AI and Human Diversity
One of the greatest limitations of AI in perfumery is human genetic diversity. Our olfactory systems contain hundreds of functional odorant receptor genes, and genetic variations influence how we perceive particular aromatic molecules. Some people are exceptionally sensitive to certain musks, while others experience partial or nearly complete insensitivity to them (Iso-E is a classic example). A floral molecule that seems delicate and beautiful to one person may smell sharp, unpleasant, or overwhelmingly intense to another.
These differences are not simply matters of preference. They can arise from biological differences in how odor molecules interact with our sensory receptors. AI fragrance models trained on population averages necessarily compress some of this variation.
A perfume optimized for the statistical average may still be unpleasant, overwhelming, or almost imperceptible to a significant number of individuals. Age, sensory adaptation, acquired smell disorders, and individual sensory histories introduce additional complexity. This raises an important question for the future of AI-generated perfume:
When AI fragrances are optimized for the average consumer, whose sense of smell becomes the standard and whose experience disappears into the margins?
2. The Human Body is Not Merely a Surface for Depositing Perfume
One of the most common questions about perfume is why the same fragrance smells different on different people. At Alkemia Perfumes we find that the answer involves a complex relationship between fragrance composition, human skin, evaporation, temperature, and perception.
Human skin contains lipids, moisture, salts, microorganisms, and compounds produced by the body. These factors can influence how fragrance ingredients evaporate, diffuse, persist, and are experienced throughout the day. A perfume that smells exquisite on a laboratory blotter may develop quite differently when worn on the wrist or throat. Some notes may disappear quickly. Others may become unexpectedly prominent. A delicate floral perfume may seem warmer and more musky, while a rich amber may become surprisingly restrained.
Not all of these differences result from chemical reactions with the skin. Physical evaporation, application conditions, and individual perception can also contribute.
AI can estimate molecular volatility and model certain ingredient interactions. But no broadly validated AI system can reliably predict exactly how every complex perfume will develop on every individual's skin over hours of wear. A perfume that performs beautifully in consumer preference testing models may never quite come alive on someone else's skin. That uncertainty is not simply a technological inconvenience. It is part of what makes perfume personal.
The fragrance that seems ordinary in its bottle may become extraordinary on an individual person.
3. Perfume is Shaped by the Body and Also by the Life that Body Lived
A perfume is not simply a molecular formula. It is a meeting between chemistry and biology, between culture and memory, between the intentions of its creator and the unpredictable experiences of the person who wears it.
An AI algorithm may predict that a particular combination of molecules could smell like roses after rain. But it cannot know what those roses will mean to someone who remembers a first kiss in a garden, a bouquet of flowers received from a lover, a childhood summer, or a place that no longer exists. AI may translate the visual language of a cultural moment into a technically accomplished fragrance, but it cannot understand the personal impact of that moment or choose to root a response in rebellion, joy, mourning, or celebration.
Osmo's Principal Odor Map represents an important scientific advance in predicting the odor characteristics of individual molecules, but a finished perfume is rarely composed of a single molecule. A fragrance may contain dozens or hundreds of aromatic constituents interacting at different concentrations. And those interactions are not always simple. Two pleasant-smelling materials may become unpleasant together. A nearly imperceptible ingredient may dramatically alter an entire composition. One aromatic material may suppress another, while a trace component may introduce a surprising new facet.
Concentration can also change perception. A material that smells delicate at low concentration may become harsh or medicinal at higher levels. This matters because perfumery is not merely the selection of ingredients with desirable individual odor profiles. Just as a symphony is more than a series of musical notes, a truly special perfume is more than single molecules.
AI may predict which fragrances will appeal to the greatest number of consumers. But it cannot determine which perfume will make one person feel unmistakably, wonderfully themselves.
4. Can AI Understand the Emotional Connection Between Perfume and Memory?
The ability to predict certain measurable aspects of smell should not be confused with the ability to understand the entire human experience of fragrance. Smell is intimately connected with memory and emotion.
A particular fragrance can evoke an experience that has been forgotten for decades, sometimes with astonishing immediacy.
The scent of incense may evoke sacred ritual for one person, a childhood home for another, and traveling to an unfamiliar environment for someone else. The smell of damp earth may suggest renewal, grief, agriculture, monsoon rains, or the forests of childhood. Even familiar fragrance ingredients such as rose, sandalwood, vanilla, and patchouli carry meanings that differ dramatically across cultures and generations.
AI systems can identify patterns in language, imagery, and recorded associations. AI may learn that nostalgia is frequently associated with powdery musks, sweet gourmands, or familiar household aromas. But identifying a statistical association is not the same as understanding a lived experience.
An algorithm cannot know whether the scent of lilacs reminds someone of a late beloved grandmother, a special childhood moment, or another entirely private experience. Nor can it comprehend the significance of a fragrance whose meaning is rooted in cultural traditions that are not represented in commercial datasets.
AI can create agreeable smells, it cannot convey the personal and cultural stories, experiences, and deeper meanings that bring true connection to fragrances.
5. Empathy, Extraction, and The Ethics of AI Driven Fragrances: Your Emotions are Marketing Segments
Perhaps the most troubling possibility in AI-assisted perfume creation is the growing relationship between cultural forecasting, emotional analysis, and commercial fragrance development.
Trend forecasting companies examine consumer behavior, emerging aesthetics, emotional attitudes, and macro-societal shifts to identify commercial opportunities. These insights raise uncomfortable ethical questions.
What happens in a brave new world where collective anxiety becomes an input for a system designed to manufacture products that capitalize on that anxiety?
Consider a society experiencing economic uncertainty, loneliness, political instability, and digital exhaustion.
Trend analysis translates this pain into desire for comfort, nostalgia, escapism, and emotional reassurance. AI-assisted fragrance platforms turn the psychological vulnerabilities of the individual consumer into perfumes and marketing strategies selling warmth, safety, familiarity, and playful relief. While such perfumes may be pretty or even genuinely comforting, the concern lies in the broader commercial possibility: what are the ethics of using unprecedented AI technological efficiency to convert collective emotional distress into marketable aesthetics?
There is an enormous difference between creating something beautiful because people are suffering and creating something profitable because people are suffering.
One begins with empathy. The other begins with extraction.
The Future of Perfumery: Artificial Intelligence vs Human Art
At Alkemia, we believe that perfume is more than a collection of aromatic molecules and that you are more than a marketing segment. Your perfume is an expression of your human individuality, an invitation into your world, and way of writing an olfactory story that becomes uniquely your own when it meets the alchemy of your skin.
For us, the purpose of perfume has never been merely to smell statistically pleasing. It is to surprise us. To move us. To awaken something forgotten. To express something for which we have no other language.
AI have a geometry for smell but it is humanity that retains the unchartable geography of longing, memory, heart, and spirit.
Scientific References and Further Reading
-
Lee, B. K., et al. (2023). A Principal Odor Map Unifies Diverse Tasks in Olfactory Perception. Science, 381, 999–1006.
- WGSN (2026). Why Multi-Sensory Experiences Are the Next Marketing Frontier.
-
Deep Learning for Odor Prediction on Aroma-Chemical Blends. Research on computational modeling of fragrance mixtures.
Art: Aeroritratto di Mussolini aviatore (Aerial Portrait of Mussolini the Aviator) by Alfredo Gauro Ambrosi, 1930