The Saffron Census: We Read Every Label in the Database. Some Aren’t Saffron.

Saffron is the most expensive spice on earth — it takes something like 150,000 hand-picked flowers to make a kilogram of the red stigmas the price is about. That makes its supplement shelf a natural experiment: what happens to label honesty when the ingredient actually costs money?

This one is different from our previous counts in a way we’re a little proud of: the category is small enough that we didn’t sample it. We read every saffron record in the NIH label database — all 226 of them. A census, not a sample.

First finding: some of the saffron isn’t saffron

Nineteen of the products that come up under “saffron” are a different plant entirely. They’re safflower — Carthamus tinctorius, a thistle relative that costs pennies — sold under its historical common name, “American saffron.”

To be precise about what this is and isn’t: the name is old, traditional, and legal, and the labels we read do say “safflower” somewhere on them. Nobody is technically lying. But these products sit in the same search results, at a glance of the front label they read as saffron, and the plant inside has none of the compounds the saffron research measures. The world’s most expensive spice comes with a legal body double, and one in twelve “saffron” records in the database is actually it.

The 159 products containing real Crocus sativus are what the rest of this article counts.

Second finding: the shelf has two favourite doses, and both are borrowed

Among capsule and tablet products stating a saffron dose (107 of them), the doses don’t spread out smoothly. They pile up at two spots:

• 28–30 mg — 25 products. This is the dose range of the modern mood-and-evening research line: the four-week sleep-quality trial used a standardized extract at 14 mg twice daily — 28 mg a day (PMID 32056539).
• Exactly 88.5 mg — 34 products. That oddly specific number isn’t random either: it’s half of 176.5 mg, the daily dose used in the appetite research on a different branded extract, Satiereal (PMID 20579522).

So far, so evidence-based — until you check who actually names a branded extract: 9.4% of products. Ten name Satiereal. Four name affron. Thirty-four products print the Satiereal number; ten contain the thing the number came from.

The dose survives; the ingredient doesn’t. A trial number migrates onto labels as a kind of borrowed credibility, detached from the specific standardized extract that earned it. We’ve seen doses hidden, inflated and diluted in this series — this is a new one: doses inherited.

Third finding: the identity gaps, again

Saffron repeats the pattern we found in ashwagandha almost number for number:

• 60% don’t say which part of the plant they used. The trials used stigma extracts — the expensive part. Only 28% say “stigma”; 13% admit petal or flower, the cheap part.
• 70% state no standardization — no crocins, no safranal, no percentage. (Ashwagandha’s number was 71%. Apparently this is where the category settles when nobody asks.)
• 30% of all real-saffron products hide the dose in a blend — 0% among dedicated saffron bottles, 49% once saffron is a supporting ingredient in someone else’s formula. The dilution effect, seventh ingredient in a row.

For a spice whose price gap between stigma and petal is enormous, “which part” and “standardized to what” aren’t details. They’re most of the price tag.

An almost-empty shelf

One number surprised us by being tiny: of 159 saffron products, only three are marketed for sleep. The sleep-quality trial is from 2020; the shelf hasn’t caught up with its own newest research line. Most saffron on the market is sold for mood, eyes, or as one name among twenty in a blend.

Where we stand — and this time the conflict is total

Our formula contains affron saffron extract at 28 mg — the branded extract and the exact dose of the trial above. This article, in other words, describes our own label as the thing the category mostly fails to be. That’s the strongest conflict of interest in this series, so here is everything that cuts against us:

• The sleep trial we cite was funded by Pharmactive, the company that makes affron. It’s a real randomized controlled trial in a peer-reviewed journal, and industry funding is disclosed in it — but you should know it when weighing the evidence, so we’re telling you.
• A 28 mg dose is not “stronger” than 88.5 mg. Different extracts, different concentrations, different research lines. Our point is not that one number wins; it’s that a number without its extract is unverifiable.
• A census of the database is not a census of the market. DSLD submissions are voluntary; marketplace-only brands may be missing entirely.
• Liquid tinctures (6 products) were excluded from dose statistics — they state herb-equivalent weights that can’t be compared to extract milligrams.

How to read a saffron label in ten seconds

1. Check the Latin name. Real saffron is Crocus sativus. “American saffron” and Carthamus tinctorius are safflower — a different plant.
2. Look for “stigma.” That’s the part the research used and the part the price is about. Sixty percent of labels won’t tell you; that silence is information.
3. Match the number to its extract. 28 mg and 88.5 mg each mean something for the standardized extract that was studied at that dose. On a label that names neither an extract nor a standardization, the same digits are just digits.

How we counted, so you can check us

We pulled every on-market record matching “saffron” in the DSLD — 226 labels, read in full on 29 August 2026. Nineteen safflower products were excluded from the saffron statistics and reported separately; 159 products list Crocus sativus material in the panel. Doses were read from the saffron line (blends walked, units converted); dose statistics use the 107 capsule/tablet products with a stated dose. Plant part, standardization and branded-extract mentions were read from the ingredient lines, product names and label statements. Both trial doses are cited from their PubMed abstracts, linked above, including the sleep trial’s funder.

The database is free and needs no account. If you repeat this and get different numbers, we’d like to know.

†These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease. This article is educational and is not medical advice; if you have a medical condition or take medication, talk to your healthcare provider.

There is a fifth mechanism beyond borrowed doses. In the apigenin census we found a dose with no trial parent at all — 50 mg, traceable to a podcast, printed on eight of nine dedicated bottles.