AI Can Now Smell a Fake

AI machine learning detecting spice fraud in supply chain

In my last post about Vikram’s cumin farm in Rajasthan, I described the problem of adulteration. Cumin cut with coriander powder. Turmeric boosted with lead chromate. Chili powder stretched with brick dust.

I told you it was systemic that the big spice companies know. That their supply chains are built for margin, not purity.

Some people pushed back. “How bad can it really be? We have food safety standards. There are regulations.”

Here’s how bad it is.

TL;DR: Spice fraud is a multi-billion dollar problem — lead chromate in turmeric, ethylene oxide in cumin, fake saffron everywhere. AI and machine learning are now detecting adulteration faster and cheaper than traditional lab methods. Here’s what’s changing.

Researchers found that between 20 and 30 percent of commercial saffron sold globally is adulterated. In some regions, that number hits 60 percent. Not 6. 60. More than half the saffron on shelves isn’t what the label says it is.

And saffron is the world’s most expensive spice. If they’ll fake the expensive stuff, what do you think happens with everyday cumin and coriander?

Turmeric, our beloved turmeric, is often adulterated to meet demand for its bright color, especially in powdered form, where additives are hard to spot visually. Because rhizomes yield limited powder, economic pressures encourage dilution. Lead chromate is commonly added to enhance color, and its use is linked to higher blood lead levels. As a result, turmeric is considered one of the most hazardous adulterated spices.

In January 2026 alone, European food safety authorities logged 189 fraud reports. Ethylene oxide in spice mixes. Undeclared fillers. Mislabeled origins. That’s one month in one region.

Unfortunately, this is the standard in the industry, not an isolated issue.

But something is changing. And it’s changing fast. And Trevean Spice is here to make it right.

Can Machines Actually Learn to Taste?

Researchers are now using machine learning, deep learning, and computer vision to verify the authenticity of spices more accurately. Instead of simply ticking off items on a compliance checklist, these methods analyze the contents of a spice jar to ensure they match the label’s claims.

The technology works across multiple methods. Near-infrared spectroscopy combined with AI models can detect adulterants in powdered foods with over 90 percent accuracy. Researchers developed deep learning models that identify cinnamon adulteration using image recognition alone, achieving 97 percent accuracy. AI systems can now perform origin identification, meaning they can tell not just what’s in a spice but where it was grown.

Read that last part again. AI can identify the geographic origin of a spice by analyzing its chemical composition.

When Rushi held up two handfuls of cumin in Vikram’s home — one from a bulk distributor, one from his farm — she could smell the difference. The distributor sample was flat and dusty. Vikram’s was warm, peppery, and complex.

That sensory gap? Machines can now measure it. Precisely. Repeatably. At scale.

The volatile oil content that gives cumin its actual flavor. The chemical markers that distinguish cumin grown in Rajasthan’s Thar Desert from cumin grown somewhere else and blended to look the same. The fillers that shouldn’t be there.

All of it is now detectable by algorithms that don’t get tired, don’t take bribes, and don’t have financial incentives to look the other way.

What Does This Mean for “Good Enough” Quality?

Here’s where this gets relevant for every product manager reading this. Not just the ones building food products.

For decades, “good enough” was protected by the gap between what companies shipped and what customers could verify. If your cumin was 70 percent cumin and 30 percent filler, the customer probably couldn’t tell. They’d never had the real thing. They didn’t know what they were missing.

That’s the photocopy-of-a-photocopy problem I described in The Cumin Test. The store-bought cumin wasn’t bad in a way you’d notice if you’d never had pure cumin. It was just less. Diluted. A shadow.

“Good enough” survived because verification was expensive, slow, and inaccessible.

That’s the end.

When AI can analyze a spice sample in seconds and tell you exactly what’s in it, where it came from, and whether it matches the label, “good enough” becomes measurable. The gap between what you ship and what you claim gets a number attached to it.

And this isn’t limited to spices.

Think about every product category where customers trust the label because they can’t independently verify it. Supplements. Cosmetics. Organic produce. “Natural” anything. Premium anything.

AI-powered authentication is coming for all of it. Portable spectroscopy devices are getting cheaper. Smartphone-based analysis tools are in development. The cost of proving what’s inside a product is dropping toward zero.

If your product is what you say it is, this is the best thing that could happen to you.

If it isn’t, well, the clock is ticking.

What’s the Science Behind What Farmers Already Know?

Let’s connect this back to Trevean Spice because the science is validating something Rushi understood instinctively.

When we test our batches, we’re looking at volatile oil content. That’s the compound that gives a spice its actual flavor and aroma. It’s the thing that separates cumin that hits you before it reaches your nose from cumin that smells like dust.

Vikram’s cumin consistently has a dramatically higher volatile oil content than store-bought alternatives. Not marginally higher, but dramatically higher.

Why? Because his cumin is pure, single-origin, harvested at the right time, and dried naturally in desert air. It hasn’t been blended with lower-grade seed from other regions. It hasn’t sat in a warehouse for months, losing potency. It hasn’t been stretched with filler to improve someone’s margin.

The AI authentication research confirms this pattern at scale. When you analyze large datasets of spice samples, the pure, single-origin products cluster together with distinct chemical profiles. The adulterated products show diluted signatures — lower volatile oil, foreign compounds, and inconsistent profiles.

Rushi could smell this in Vikram’s home. The machines can now prove it in a lab.

And soon they’ll prove it at the point of sale. That is the intersection we are playing in.

Why Does AI Fraud Detection Change the Product Management Conversation?

If you’re a PM, I want you to think about this through the lens of your own product.

Every product has inputs: raw materials, components, data sources, APIs, and content. Something goes in, you transform it, something comes out, pretty simple math.

The quality of what goes in determines the quality of what comes out. Everyone knows this. Nobody argues with it.

But here’s what most PMs actually do: they optimize the transformation. Better processes. Better tooling. Better systems. They spend 90 percent of their energy on what happens in the middle and 10 percent on what goes in at the beginning.

This is the equivalent of a spice company investing in better packaging while buying adulterated cumin.

The cumin test isn’t about cumin. It’s about verifying your inputs before you build on top of them.

In software, this means: do you actually know the quality of the data feeding your product? Or do you trust the pipeline because it’s always worked?

In services, this means: do you know the real capability of the people delivering your product? Or do you trust the resume?

In physical products, this means: do you know what’s actually in your raw materials? Or do you trust the supplier because the price is right?

AI is making input verification cheap and fast across every domain. The PMs who get ahead of this will be the ones who verify first and build second.

The ones who don’t will ship products that look right on the surface but fall apart under inspection. Just like cumin, which smells fine until you put it next to the real thing.

How Does This Connect to the Product Onion Framework?

If you’ve been following my Product Onion series, this maps directly to the core layer.

The Product Onion says you build from the inside out. The innermost layer — the core — is the fundamental thing your product does. The truth of what it is. Everything else (features, packaging, marketing, brand) wraps around that core.

If the core is compromised, no amount of outer layers will save you.

Adulterated cumin with beautiful packaging is still adulterated cumin. A software product with a gorgeous UI built on bad data is still a bad product. A service with a great sales pitch delivered by the wrong people is still a bad service.

What AI authentication does is strip away the outer layers and examine the core directly. It skips the packaging. It ignores the brand. It looks at the substance.

And increasingly, customers will have access to this kind of verification. Maybe not through a spectrometer in their pocket (not yet). But through lab reports linked to NFC tags on packaging. Through third-party verification platforms. Through AI tools that aggregate and analyze product data.

The outer layers of the Product Onion are getting thinner. The core is getting more exposed.

Build accordingly.

What Is Trevean Spice Doing About Spice Fraud?

We already test every batch. That’s been our practice since the beginning.

But the emergence of AI-powered authentication is changing what we can do with those test results.

Right now, our batch testing proves the absence of adulterants and measures potency. That’s defensive. It tells you what’s not in the jar.

What AI authentication makes possible is something more ambitious: a verifiable chemical fingerprint that links a specific jar of Rajasthan Gold back to a specific harvest from specific farms.

Not “sourced from India.” Not even “sourced from Rajasthan.”

Sourced from Vikram’s field in Barmer district, harvested in the third week of March, dried for 12 days in natural desert air, blended with coriander from the neighboring farm and black cardamom from the eastern hills.

That level of specificity used to be impossible to verify at scale. It required trust. You had to believe the label because there was no way to confirm it independently.

AI changes that equation. The chemical signature of Vikram’s cumin is distinct. Measurably distinct. And as authentication tools become more accessible, we’ll be able to link our test results to our origin stories in a way that’s not just claimed but proven.

This is where transparency stops being a marketing message and starts being a product feature.

What’s the Honest Truth About Fighting Spice Fraud?

I want to be direct about something.

We’re not there yet. The AI authentication tools that exist today are mostly in research labs and large-scale industrial applications. Portable devices capable of real-time spice authentication at the consumer level are in development but not widely available.

So when I say “AI can now smell a fake,” I mean the technology works. The science is proven. The accuracy is there. What’s still catching up is the accessibility and cost for small companies like ours.

But the trajectory is clear. And if you’re building a product today, you should build it assuming that everything you ship will eventually be independently verifiable.

Not because you’re afraid of getting caught.

Because if what you’re shipping is real, verification is the best marketing you’ll ever have.

What Should You Do on Monday Morning?

Here’s the practical part.

Audit your inputs. Pick the most important raw material or component in your product. When was the last time you independently verified its quality? Not through the supplier’s own documentation. Through your own testing or a third-party analysis. If the answer is “never” or “it’s been a while,” that’s your first action item.

Quantify the gap. If you can, get a sample of your input tested against the best available version of that input. What’s the difference? Is your cumin a 3 out of 10 on volatile oil content, while the single-origin version is a 9? Is your data source 60 percent accurate when you assumed 95 percent? Know the number because someone else will measure it eventually.

Document what you find. Whether the results are great or embarrassing, write them down. If they’re great, that’s a product story. If they’re embarrassing, that’s a product decision you need to make. Either way, you’re better off knowing.

Follow the verification technology in your industry. AI-powered authentication is expanding into every product category. Know what’s coming. Know when portable verification will reach your customers. Build your product to welcome that scrutiny, not fear it.

The era of unverifiable “good enough” is ending.

AI isn’t just finding fakes. It’s creating a world where the distance between what you claim and what you ship is precisely measurable.

For companies that have been cutting corners, this is a reckoning.

For companies that have been doing the work, this is proof.

I know which side of that line I want Trevean Spice to be on. You should figure out where your product stands before the machines do it for you.

This is the second post in a series that started with “The Cumin Test” and continued with “Your Spice Has a Passport Now.” Same story. Same farm. Each post pulls a different product management lesson from the same decision: know where things come from, or get left behind by the people who do.

Related Reading

Follow The Product Manager’s Journal or subscribe to the blog for the next one.

Frequently Asked Questions

How common is spice fraud and adulteration?

Extremely common. Studies show 20–30% of commercial saffron is adulterated. Turmeric is frequently boosted with lead chromate for color. Cumin is cut with coriander or other fillers. The spice trade’s complexity and limited testing make fraud profitable and hard to detect.

How does AI detect food fraud?

AI uses spectroscopy data, chemical fingerprints, and pattern recognition to identify adulterants that traditional lab tests might miss. Machine learning models trained on thousands of samples can detect anomalies in seconds — faster and cheaper than sending every batch to a lab.

What contaminants are commonly found in spices?

Common contaminants include lead chromate in turmeric (for color), brick dust in chili powder (for weight), coriander powder in cumin (cheaper filler), ethylene oxide as a fumigant, and artificial dyes in saffron. Some of these pose serious health risks including cancer and organ damage.

How can consumers verify their spices are authentic?

Look for brands that offer supply chain transparency — NFC tags, QR codes, or digital product passports that let you trace the spice to its source. Buy from companies that name their farmers and share test results. If a spice is significantly cheaper than market rate, that’s a red flag.