You're standing in front of two bottles. One says 600 mg of ashwagandha. The other says 500 mg and costs less. The obvious move is to take the bigger number, and it's the wrong move — because on most labels that number isn't comparable to anything, including the number on the bottle next to it.
We can say that with some confidence, because we went and counted. The National Institutes of Health runs a public database of dietary supplement labels called the DSLD. Anyone can search it. We pulled 336 ashwagandha labels that are currently on the market and read every one.
We expected underdosing. We were wrong.
Going in, our assumption was the one you hear everywhere: that most supplements contain a fraction of what the research used. For ashwagandha the clinical range is roughly 300–600 mg a day.
The data didn't support it. Of the products that state a dose, only 15% fall below 300 mg. Nearly half — 49% — sit at 600 mg or above. On raw milligrams, the shelf is in better shape than the internet suggests, and we'd rather say that plainly than keep repeating a claim our own numbers contradict.
The real problem turned out to be somewhere else, and it's worse.
Seven labels in ten can't be compared to any study
Ashwagandha is sold as an extract, and extracts are concentrated. How concentrated is described by the percentage of withanolides — the compounds the research measures. In our sample, the disclosed figures ran from 0.5% to 35%.
Sit with that range for a second. Six hundred milligrams of a 1% extract and six hundred milligrams of a 5% extract are not the same product, and they're not off by a little. They're off by five times. Same number on the front. Same shelf. Different products.
Here's what we found across those 336 labels:
- 29% state how standardized the extract is.
- 71% don't state it at all — their dose cannot be compared to a study, or to a competitor, or to anything.
- Of the ones that do disclose: 47% are weaker than 5% withanolides.
- 27% don't say which part of the plant was used.
- 18% name a specific branded extract.
336 ashwagandha labels · NIH database
Does the label say how concentrated it is?
Each square is one product. Whether it states how concentrated its extract is.
Disclosed potency, where stated
Six hundred milligrams of a 1% extract and 600 mg of a 5% extract are not the same product. They are off by five times. Same number on the front, same shelf, different material.
None of this is illegal. A label saying "Ashwagandha 600 mg" is not making a false statement. It's making an unverifiable one — and that distinction is invisible to almost everyone standing in the aisle.
Root or leaf changes the answer too
The clinical trials behind ashwagandha were run on root extract. Leaf contains withanolides as well — often more of them, which makes a leaf or leaf-and-root blend look impressive on a certificate of analysis while being a different material from the one that was studied.
More than a quarter of the labels we read don't say which part they used. We wrote a longer breakdown of root versus leaf, because it's the single easiest thing to get wrong when comparing two bottles.
How to read an ashwagandha label in ten seconds
Three things have to be present. If any one of them is missing, the other two can't be interpreted.
- The plant part. Look for "root" or "root extract". If it says only "ashwagandha", you don't know what you're buying.
- The standardization. A percentage of withanolides, stated as a number. No number means no comparison.
- The dose. Only now does the milligram figure mean something — and only against a study that used the same plant part at the same concentration.
A named branded extract is a shortcut through all three, because branded materials publish their specification. That's not an argument for any particular brand — it's an argument for knowing what you're holding.
How we did this, so you can check us
We searched the DSLD for "ashwagandha", filtered to products currently on the market, and read 336 labels in full on 28 August 2026. The database held 3,772 matching records in total, so this is a sample, not the whole shelf.
For each label we took the ashwagandha line from the supplement facts panel, converted every unit to milligrams, and searched the full label text for a stated withanolide percentage and for the plant part. Fifty-seven labels gave no dose in milligrams; they were excluded from the dose numbers and kept in the transparency numbers. Nothing else was excluded.
The database is free and requires no account. If you repeat this and get different numbers, we'd genuinely like to know.
Why we bothered
We make a sleep supplement, so treat this with the suspicion it deserves. But the reason we counted is that we had to answer the same question for ourselves before we printed anything: what does a dose have to look like to be worth stating?
Our answer was that all three facts go on the label or the ingredient doesn't go in — the part of the plant, the standardization, and the milligrams. You can see every dose next to the study it came from in the Evidence Library, along with the research we decided not to cite and why, and the twelve ingredients that didn't make it into the formula at all.
The shelf isn't full of cheats. It's full of numbers nobody can check. Those aren't the same problem, and only one of them is fixable by reading more carefully.
†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.
We ran the same count on a second category. 588 melatonin labels turned out to have the opposite problem: the doses are perfectly clear, and almost none of them match the dose the research used.
There is a sequel to this count. When ashwagandha appears inside a multi-ingredient sleep formula, its median stated dose falls to 125 mg — a quarter of the standalone median, below every trial.
The pattern repeated almost number for number when we later ran a full census of saffron labels: 70% state no standardization, 60% no plant part — and one in twelve “saffron” records is a different plant altogether.
One more count worth knowing: across 357 ashwagandha labels in our warnings census, zero mention autoimmune conditions — while “immune” appears 25 times as a selling point.