The Internet of Smells: Translating Air into Data and Memories into Formulas

The promise of the Internet of Senses (IoS) is to expand the digital world beyond audiovisual, introducing the remote transmission of smells. However, digital s

The expansion of the digital ecosystem toward the Internet of Senses (IoS) marks the overcoming of the purely audiovisual hegemony that has dominated the information age. The engineering promise is that of an immersive multisensory telepresence, a technological horizon closely tied to the development of 6G networks and shared virtual environments. However, as far as smell is concerned, technology is taking two separate and asynchronous steps: on the one hand, detecting and classifying an odor through chemical sensors and artificial intelligence; on the other, reproducing it remotely through mechatronic devices that release mixtures of molecules. If the first phase has reached a degree of industrial maturity, the second remains dramatically limited, expensive, and far from the utopia of a universal, pocket-sized "scent transmission."

The Vision and the Prototype: From Text to Molecule

The conceptual framework of digital scent technology has been formalized in academia as the engineering capability to detect, transmit, and receive digital olfactory media. As highlighted by the IEEE (2022), the goal is not the banal shipping of a digitized scent, but the synchronized integration of smell, sound, image, and haptic feedback to generate remote experiences indistinguishable from physical reality [2981]. The literature anticipates that by 2030 the infrastructure of edge intelligence and semantic communications will make it possible to manage olfactory and gustatory streams in real time [2984].

This is not pure theoretical speculation. A formidable example of the current state of the art is AromaGen (arXiv, 2026), a wearable hardware device designed to generate interactive olfactory experiences. The system translates textual inputs, images, or voice commands into a mathematical vector of proportions based on 12 pre-loaded base odorants, releasing them through a miniaturized dispenser. The authors of AromaGen have even successfully demonstrated the feasibility of a rudimentary remote transmission of an olfactory experience between two different cities [2992]. Yet the intrinsic limitation of this approach is obvious: the entire reproducible olfactory universe is forcibly constrained to an extremely limited chemical palette of only 12 base ingredients.

Reading the Air: Electronic Noses and Digital Fingerprints

The process of digitizing an odor begins with reading it. Electronic noses (e-noses) integrated with artificial intelligence have made it possible to create genuine "fingerprints" of odors. A review in Artificial Intelligence Review (2025) cites systems capable of discriminating with 78% accuracy distinct macro-olfactory categories, such as clean air, onion fumes, or gasoline vapors, using machine learning to classify the chemical reaction patterns of the sensors [2980].

The most mature commercial application of these systems concerns quality control and food safety. As documented in PMC (2025), many industrial e-noses achieve accuracies above 90% in classifying well-defined food classes (for example, detecting the degree of ripeness of fruit or the onset of decomposition in meat) [2986].

However, the sensing hardware bears insurmountable chemical-physical limitations. An analysis in Intelligent Computing (2023) dissects the vulnerabilities of electronic noses: partial selectivity, low sensitivity to minimal concentrations, extreme instability over time (the so-called sensor drift), poor reproducibility, and susceptibility to environmental noise [2990, 2993]. Moreover, an odor in the real world is a chaotic mixture of hundreds of volatile compounds in continuous mutation, not an audio file with a clear waveform to be sampled, compressed, and sent. As emphasized in Trends in Food Science & Technology (2024), predictive models work excellently on standardized laboratory datasets, but struggle enormously with complex mixtures. Above all, predicting the chemical data of a gas is in no way equivalent to predicting the subjective perception that a specific person will have of that gas [2985].

The Problem of Reproduction: Why the "Scent Modem" Is Missing

If the sensing phase (detection) is advancing rapidly, the actuation phase — the physical reproduction of the odor — represents the true bottleneck of the Internet of Senses. Synthesizing sounds or images requires the manipulation of electrons and photons through speakers and screens, elements that are physically clean and rapidly decaying. Synthesizing an odor requires the physical delivery of matter (volatile molecules) into the user's space, matter that deposits, mixes, persists, and saturates the environment.

Some laboratory successes demonstrate that targeted reproduction is chemically possible. Researchers at Tokyo Tech (2022) extracted 20 primary odor components from 185 essential oils through mass spectrometry and multidimensional analysis. Using a multicomponent olfactory display, they managed to reproduce seven specific essential oils with such precision that, in subsequent double-blind sensory tests, no statistically significant difference emerged between the original odor and the artificially reproduced one [2991].

The cost of this success, however, is extreme hardware complexity. The Japanese Journal of Applied Physics (2026) illustrates the requirements of these olfactory displays: they require mass flow controllers, solenoid micro-valves, autosamplers, and complex fluid dynamics systems to mix and push the air [2989]. This infrastructure perfectly explains why the "scent modem" is not yet a pocket-sized object to be integrated into the chassis of a smartphone: the mechanical bulk and the need to constantly recharge chemical cartridges make the device's scalability a logistical nightmare.

The Value and Privacy of Olfactory Memory

Net of the engineering obstacles, the evolution of digital scent technology raises entirely unprecedented editorial, ethical, and regulatory questions. The real question is not the frivolous and overhyped "when will we receive the scent of a coffee via WhatsApp?", but a question of power and control: who holds the authority to decide which odors deserve to be digitized, archived on servers, and subsequently resold?

Smell is the human sense most intimately, biologically linked to the limbic system, the neurological seat of long-term emotional memory. An odor does not simply transmit information; it instantly evokes a childhood home, the intimacy of a specific person, the identity of a place, or the memory of a trauma. Transforming this primordial sense into binary data and a commercially reproducible chemical mixture raises questions of privacy, cultural appropriation, and personal consent that audio and video have never made so urgent [2981, 2984, 2992]. If a wearable or environmental sensor "smells" and records the chemical profile of our home, our breath, or our skin, it is acquiring biometric and behavioral data of unheard-of depth.

The internet of smells will not simply be a picturesque network transporting scents from one continent to another. It will, to all intents and purposes, be a surveillance and stimulation infrastructure that transforms the air we breathe into extractable data, our most intimate memories into patentable chemical formulas, and our involuntary emotional reactions into consumer products. The true legal and human challenge of the coming decades will not be so much succeeding in transmitting a perfectly synthesized odor kilometers away, but establishing whether we will be able to defend, in a hyper-connected and multisensory ecosystem, the fundamental right to no longer want to smell.

Bibliographic References and Sources

  • [Digital Scent Technology: Toward the Internet of Senses — IEEE, 2022] [2981]
  • [The Internet of Senses: Building on Semantic Communications and Edge Intelligence — arXiv, 2022] [2984]
  • [AromaGen: Interactive Generation of Rich Olfactory Experiences — arXiv, 2026] [2992]
  • [Advances in artificial intelligence for olfaction and gustation — Artificial Intelligence Review, 2025] [2980]
  • [Review on Algorithm Design in Electronic Noses — Intelligent Computing, 2023] [2990, 2993]
  • [Bridging odorants and olfactory perception through machine learning — Trends in Food Science & Technology, 2024] [2985]
  • [Scientists Artificially Recreate Smells Using an Olfactory Display — Tokyo Tech, 2022] [2991]
  • [Digital scent technologies and their applications — Japanese Journal of Applied Physics, 2026] [2989]
  • [A Comprehensive Review on Sensor-Based Electronic Nose for Food Quality and Safety — PMC, 2025] [2986]

Article by the Editorial Staff of La Bussola dell’IA