Here is an experiment you can run in any supplement aisle. Pick up a bottle of plain ashwagandha. The label will say something like 500 or 600 mg. Now pick up a “sleep formula” that lists ashwagandha on the front. Find its dose on the back — if it’s printed at all.
We ran that experiment across the whole shelf, using the NIH’s public label database. In our previous piece we counted how often doses go missing in 600 sleep formulas — and admitted the count had a blind spot: a stated dose isn’t automatically a meaningful one. This is the follow-up. We took every stated dose and checked it against the dose the trials actually used.
The same ingredient shrinks when it joins a formula
Across the standalone ashwagandha products we read in our first study — 279 with a stated dose — the median is 500 mg.
Inside the 600 sleep formulas, 73 products list ashwagandha as an ingredient. Then it goes like this:
- 41 of the 73 don’t state its dose at all — it sits inside a blend.
- Among the 32 that do, the median is 125 mg — a quarter of what the standalone bottles carry, and well under the 300 mg where the clinical trials start.
- Exactly one of the 32 reaches 300 mg. And that one lists ashwagandha powder — raw ground root, not the concentrated extract the studies used. On like-for-like material, the count is zero.
So of 73 sleep formulas advertising ashwagandha, the number whose stated dose matches the research on the material the research used is 0.
It isn’t just ashwagandha
We ran the same check on every ingredient where we could verify the studied dose against a specific trial — not “commonly cited” numbers, but doses we could pin to a PMID and read.
| Ingredient | Dose the trial used | Median on the shelf | Stated doses reaching it |
|---|---|---|---|
| Glycine | 3,000 mg (PMID 22529837) | 50 mg | 0 of 15 |
| Ashwagandha | 300–600 mg extract | 125 mg | 1 of 32 — and it’s powder |
| Valerian | 400 mg (PMID 7122669) | 200 mg | 16 of 125 (12.8%) |
| L-theanine | 200 mg (PMID 31623400) | 100 mg | 83 of 193 (43.0%) |
Glycine is the extreme case. The trials behind it used three grams — a heaped dose that takes real capsule space. The median glycine line in a sleep formula is 50 mg: sixty times less. At that level it isn’t an ingredient. It’s a word on the label.
Put together, across these four ingredients: 365 stated doses, and 100 of them — 27.4% — reach the dose a trial used. Take out L-theanine, the one bright spot, and it drops to 9.9%.
The dilution effect · NIH database
Ashwagandha: alone vs. inside a sleep formula
Median stated dose per serving. Same ingredient, same shelf.
↓ a four-fold drop — and 41 of 73 formulas don’t state the dose at all
Stated doses that reach the trial dose
The exception: melatonin doesn’t shrink. Median 3 mg in near-single-ingredient products, 3 mg in the largest formulas. The expensive actives dilute; the cheap hormone keeps its round number.
The exception that explains the rule
One ingredient in our data refuses to shrink: melatonin. Its median dose is 3 mg in near-single-ingredient products and 3 mg in the biggest multi-ingredient formulas. Identical. Everything else gets diluted; melatonin doesn’t.
The doses themselves are a separate problem — we counted those too, and 3 mg is ten times the physiological dose. But the fact that this one number survives formulation while ashwagandha’s collapses is telling.
Our reading — and we flag it as interpretation, not something the data proves: melatonin milligrams cost almost nothing. Botanical extracts and amino acids at trial doses cost real money and occupy real capsule volume. When a formula has twenty names to fit, the expensive ones shrink and the cheap hormone keeps its round, marketable number. The ingredient list is optimized for the front of the label, not for the studies behind it.
Add this to what we found last time
Our previous count showed that the longer a formula’s ingredient list, the more of its doses vanish into blends — from 14% hidden in short formulas to 65% in the longest. This study adds the second half: even the doses that survive in print mostly don’t reach the research.
A twenty-ingredient sleep formula now reads like this: a third of its doses aren’t stated, and of the ones that are, perhaps one in ten matches a trial. The list is long precisely because each line is cheap — at real doses, twenty ingredients wouldn’t fit in a capsule you could swallow.
Where we stand, and where this data is weak
The disclosure, as always, first: we make a four-ingredient sleep supplement whose label carries 600 mg of ashwagandha extract and 200 mg of L-theanine — the trial doses. A study finding that formulas underdose exactly these ingredients describes our product’s selling point. Weigh everything above accordingly.
And the limits are real:
- The ashwagandha comparison rests on 32 stated doses. Small sample; the direction is unambiguous, the exact median is not sacred.
- We counted per serving, as printed in the supplement facts panel. A product suggesting two servings nightly would double its numbers — and its capsule count.
- Reaching a trial dose doesn’t make a formula effective, and missing it doesn’t prove the product does nothing. Lower doses are simply territory the trials didn’t map.
- L-theanine deserves its credit: 43% of formulas that state its dose meet the studied 200 mg. Where an ingredient is cheap enough, the category can dose it honestly.
The one question that survives all four studies
We’ve now read 1,500+ labels across four counts, and every finding compresses into the same buying rule: take the ingredient you’re actually buying the product for, find its number, and compare it to the standalone bottle and the study — not to the rest of the front label.
If the number isn’t there, the formula has answered your question. If it’s there and it’s a fifth of the standalone dose, it has answered it too.
How we counted, so you can check us
We used the 600 sleep-formula labels from our previous study (NIH DSLD, products currently on market, read 29 August 2026) and extracted the stated per-serving dose for every occurrence of 14 common sleep ingredients, walking blends and converting units to milligrams. Standalone ashwagandha medians come from our 336-label ashwagandha dataset. Trial doses were verified against the linked PubMed records, not quoted from memory. Ingredients whose studied dose we could not pin to a specific readable trial — chamomile, lemon balm, passionflower, hops, tart cherry, 5-HTP, GABA, tryptophan — were left out of the threshold test and carry no verdict here.
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.
Magnesium later gave us the same picture from another angle: median 250 mg in dedicated products, 100 mg once it sits inside a sleep formula.
Theanine, the bright spot of this count, later got its own study: standalone bottles carry the trial dose almost universally, but the moment it joins any formula — sleep, stress or focus — the median halves to 100 mg.
Saffron completed the set: its census found doses hidden in 49% of the formulas that carry it as a supporting ingredient — and a new failure mode entirely: trial doses borrowed from branded extracts most bottles don’t contain.
Valerian became the eighth replication — and the strangest: its labels mix extract weight with raw-herb equivalents, so the dose question itself stops being answerable.
GABA became the ninth replication, and the steepest since ashwagandha: median 500 mg standalone, 100 mg inside formulas.