Digital Scent Technology: What Really Travels Online

Legacy & Insights

LEGACY & INSIGHTS: SCIENCE & TECHNOLOGY

Digital Scent Technology Can Share Smells. Here’s What Really Travels.

Digital scent technology can name a smell in one place and recreate a close match in another. How it does that is the story.

By Pamela F. Nichols·October 26, 2026·4 min read

Digital scent technology concept showing an electronic nose, fragrance samples, and wearable scent-delivery devices in a research laboratory. — Conceptual illustration of digital scent technology, showing how electronic sensors and wearable devices could support remote aroma experiences. Not an image of the actual Scentronic equipment.

Digital scent technology just took a real step toward something inventors have long promised: smelling something that is not in the room. A team at Ohio State University and Zhejiang Laboratory reports a system called Scentronic in Science Advances. First, one device reads a scent. Then another device, somewhere else, mists a match. Still, what crosses the network is smaller than a smell, and that is where the engineering gets interesting.

How Digital Scent Technology Reads a Smell

The idea is simple to state. A sensor sniffs the air, software names what it found, and a second device recreates something close. Scentronic starts with an electronic nose built on six metal-oxide gas sensors, according to coverage of the paper. A classifier then sorts each sniff into one of five scents: beer, cake, coffee, oranges or red wine. That is a small menu, because each scent has to be learned first. However, Reuters reports the sensor unit is portable but not wearable, at roughly 15 centimeters across.

What Actually Travels in Digital Scent Technology

Because the system sends only a label, the smell itself stays behind. The label reportedly crosses the network over Wi-Fi and a cloud server. Reported network delay was about 65 milliseconds, within a range of 57 to 76. That figure covers the message alone. Meanwhile, a separate demonstration reportedly ran from sensor to spray in about 0.6 seconds, and classification itself can take seconds. These are three different measurements. Therefore, 65 milliseconds is not the speed of smelling, and no headline should say so.

What Volunteers Could Actually Smell

Do people notice the match? In tests with 14 volunteers, people identified single scents correctly 72.86% of the time, according to reports on the paper. Next, pairs of scents reached 71.43%, against a 20% chance level, and ranking intensity reached 58.57%. However, some coverage cites up to 16 volunteers, and we have not confirmed whether that was a separate test. Overall, those scores are strong for a prototype. Still, they mean roughly one attempt in four missed, and judging intensity fell lower still.

Why Bottles Limit Digital Scent Technology

The receiving devices mist preloaded scented liquids with piezoelectric atomizers, and pulse width sets the intensity. Researcher Yang told Reuters the receiver cannot recreate every smell, because each new scent needs fresh training data and a new liquid. Likewise, the paper makes the same point. Odor is classified “at the category level and reconstructed through preloaded odor channels, rather than reproduced with full chemical compositional precision,” as quoted by TechXplore. In short, the bottle sets the limit, not the bandwidth. Because the liquids are loaded in advance, the receiver can only offer what its owner stocked. Instead of rebuilding a smell from chemistry, it recalls one.

WATCH / ABC NEWS REPORT

ABC News looks at Scentronic, the prototype that can recognize five scents. The video’s caption lists “orange wine”; the researchers’ fifth scent is oranges. Source: ABC News.

FROM KMOB1003

Where Digital Scent Technology Could Go Next

Beyond the sensor, the team built receivers as a wristband, a necklace and a handheld device, plus a cup that pairs five scents with five tastes, according to coverage. Healthcare is among the proposed uses, but it is a proposal, not a demonstration. Meanwhile, Reuters describes the work as a research prototype with no company involved. So the wearable story is about form, not proof. Moreover, none of the wearable receivers has been tested outside the authors’ lab in anything we reviewed.

What Digital Scent Technology Still Has to Prove

But the headline numbers will not settle that. Independent testing will. For that, researchers outside the team would need to show that the five-scent menu can grow without losing accuracy, and that receivers can build scents from a wider palette than bottled matches. Until then, this is a clever label system with real engineering behind it. Therefore, the safest reading is a narrow one: a working prototype that proves a label can travel, not a machine that delivers a smell. Network speed was never the hard part. The bottle is.

Signal Breakdown

Not every claim about digital scent technology carries the same weight.

Widely Reported

The paper. Researchers at Ohio State University and Zhejiang Laboratory published Scentronic in Science Advances.
The menu. Six metal-oxide sensors classify one of five scents: beer, cake, coffee, oranges or red wine.
The human tests. Coverage reports 14 volunteers scoring 72.86%, 71.43% and 58.57% on three tasks.
The receivers. Wristband, necklace and handheld devices mist preloaded liquids with piezoelectric atomizers.
The status. Reuters describes a research prototype with no company involved.

Attributed Claims and Analysis

Speed. About 65 milliseconds is reported for the network alone, with about 0.6 seconds sensor to spray in a demonstration.
The limit. Researcher Yang told Reuters the receiver cannot recreate every smell without new data and a new liquid.
The paper’s wording. Odor is classified at the category level, per a quote carried by TechXplore.
Uses. Healthcare is proposed, not demonstrated.

Not Established

whether the 14-volunteer and up-to-16-volunteer figures come from different experiments, which the paper’s methods would settle
the paper’s author order and its exact publication date, which sources give as October 21 or 22
endurance and repeated-split accuracy figures that appear in only one source
how any receiver performs outside the authors’ lab
any independent replication

Sourcing: this report draws on coverage of the Science Advances paper from Reuters, AAAS, CNET and TechXplore, as surfaced in search results. science.org and PubMed Central blocked our access, so we did not open the paper itself, and every figure above is attributed to that coverage. We found no independent testing outside the authors’ lab in the sources we reviewed. The Signal Breakdown above separates widely reported records from attributed claims and open questions.

SCENTRONIC / KEY FACTS

STUDYScience Advances paperOhio State University and Zhejiang Laboratory, DOI 10.1126/sciadv.aee9266. We have not opened the paper.

SYSTEMScentronicSix metal-oxide sensors sort a smell into one of five scents, then a label is sent to a receiver.

TESTS14 volunteers72.86% correct on single scents, 71.43% on pairs and 58.57% on intensity, per coverage.

SPEED65 ms is the networkAbout 0.6 seconds from sensor to spray in a demo. Classification can take seconds.

RADAR / WHAT TO WATCH

WEARABLESNUS light-up garmentNational University of Singapore researchers report a garment-integrated LED display in Nature (DOI 10.1038/s41586-026-11128-9).

MUSICFontaines D.C. on AI videoAfter fans objected to AI credits on the “Marianne” video, Grian Chatten told the Los Angeles Times the band will bar AI from its video contracts. That is a stated plan, not a confirmed policy.

LIVEJay-Z at SoFiJay-Z performed at SoFi Stadium on October 23 and 24. No attendance figure has been confirmed.

STREAMINGSpotify’s AI cleanupSpotify says it removed 75M+ spam tracks and took 50M+ out of recommendations. Those are separate actions, so the totals are not added.