Report 033 · Supplements
A 2026 meta-analysis pooled 22 trials and reported that ashwagandha improves stress, anxiety, and depression. The effect sizes it published are roughly ten times larger than any antidepressant ever measured. That number is the story.
Disclosure first: I formulate and sell supplements. My company makes functional-mushroom products, and ashwagandha is a root, not a mushroom, so I don't sell the thing this report is about. But I have a commercial stake in the category, which is exactly why I read the category's good news this hard. If I want a skeptic to check my shelf, I should check everyone else's the same way.
Ashwagandha (Withania somnifera) is probably the most-sold adaptogen on earth right now. It's in the stress gummies, the sleep stacks, the cortisol powders. And unlike a lot of what gets sold that way, it has a real clinical literature behind it, dozens of randomized controlled trials. So when a fresh meta-analysis pooled that literature in 2026 and came back positive, that should be good news. It mostly is. But the number it published is worth stopping on, and stopping on it teaches something more useful than the herb does.
In Complementary Therapies in Medicine, Alsanie, Alhodieb, and Askarpour ran a systematic review and dose-response meta-analysis of ashwagandha and adult mental health. Twenty-two randomized controlled trials met the criteria. The protocol was pre-registered with PROSPERO, which is the right way to do this. Their result, verbatim: supplementation "significantly improves stress (SMD = -5.88; 95 % CI: -8.15 to -3.60), depression (SMD = -5.68; 95 % CI: -8.43 to -2.94), and anxiety (SMD = -6.87; 95 % CI: -8.77 to -4.97)."
Every one of those is statistically significant, and the direction is consistent across three outcomes. If you stop reading at "significantly improves," you have a headline. The interesting part is the size.
SMD stands for standardized mean difference, and the paper says plainly that it is what they calculated: "Standardized mean differences (SMDs) were calculated." Standardizing means dividing the difference between groups by the spread of the data, so the answer comes out in standard deviations rather than in points on whatever questionnaire the trial happened to use. That's the whole purpose of it. Trials measure stress with different instruments, so you convert everything into a common currency before you pool.
That common currency has a well-worn scale attached. By the conventions almost every field uses, 0.2 is a small effect, 0.5 is medium, 0.8 is large. In mental health specifically, the UK's NICE has treated a standardized difference of 0.5 as the threshold for a clinically meaningful benefit. And for the most-studied psychiatric drugs we have, the numbers land in that same neighborhood: reviews of the antidepressant trial base, including the unpublished negative ones, put the drug-versus-placebo effect somewhere around 0.31 to 0.61 depending on which studies you include.
So hold that next to the ashwagandha result. An SMD of 6.87 for anxiety would mean the average person taking the herb ended up nearly seven standard deviations better off than the average person taking placebo. Not seven points. Seven standard deviations. That would not be a good supplement result. It would be the largest effect in the history of mental-health treatment research, beating every antidepressant, every therapy, every intervention ever measured, by more than a factor of ten, from a root extract, in trials small enough to fit into 22 studies.
When a number is that far outside the plausible range, the useful move is not to celebrate it or to dismiss it. It's to ask what units it's probably in. So look at what other people pooling roughly the same literature report. A separate systematic review and meta-analysis of ashwagandha, cortisol, stress, and anxiety, covering 15 randomized trials and 873 adults, published its results as mean differences on the actual clinical scales: anxiety on the Hamilton scale at eight weeks came out at -3.52 (95% CI -6.00 to -1.04), and perceived stress at -4.88 (95% CI -7.84 to -1.91).
Notice that those are numerically similar to the "SMD" figures in the first paper, and they are exactly the size you would expect a stress questionnaire to move. The Perceived Stress Scale runs 0 to 40. Dropping five points on it is a real, ordinary, believable result. Dropping five standard deviations is not a thing that happens.
I want to be careful here, because being careful is the point. The full text of the 2026 paper sits behind a paywall, so I have read its abstract, not its calculations, and I am not going to tell you I know which step produced those figures. What I can tell you is that the reported magnitudes are not compatible with the label attached to them, and that the authors themselves close by calling for "well-designed, high-quality trials" to "address existing heterogeneity." That is a research team flagging that their inputs were noisy. A reader should take the hint.
Probably, modestly, for stress and anxiety. That is the honest answer, and it is a genuinely decent answer for a supplement. Two independent pooling efforts, using different methods on an overlapping trial base, both found benefit in the same direction. Direction is the more trustworthy half of a meta-analysis. Magnitude is the fragile half, because it inherits every scaling decision, every small trial with an unusually tight variance, and every difference between one company's branded extract and another's.
What the evidence does not support is the version on the label: ashwagandha as a dramatic, reliable fix for how you feel. The size of the published effect is not evidence that the herb is extraordinary. It is evidence that the pooled number needs a closer look than the abstract gives it.
Effect sizes are where supplement marketing does its quietest work, because almost nobody checks them and they look authoritative. Two questions handle most of it. First, what are the units? A "mean difference" is points on some scale, and points are meaningless until you know the scale. A "standardized mean difference" is standard deviations, and there you have a fixed yardstick: 0.2 small, 0.5 medium, 0.8 large, and anything past about 1.5 in behavioral research should make you read twice, not cheer. Second, is this number physically plausible? If a supplement's reported effect beats the best-studied prescription drug in its category by an order of magnitude, the most likely explanation is not that you have found a miracle.
That's the same reflex as reading the creatine meta-analysis, where the problem was how the data were pooled rather than how big the answer came out, and the same reflex behind any "clinically proven" claim: find the actual number, then ask what it is a number of. Ashwagandha almost certainly helps some people relax a little. It is not seven standard deviations of anything.
Disclosure, again, plainly: I founded and run Shroombiosis (a company I run), which formulates and sells functional-mushroom supplements. It does not sell ashwagandha, but I have a commercial stake in the supplement category generally, which is why I hold its good news to this standard. Nothing here is sponsored and no link earns a commission; here's the full policy. A recommendation with no stake at all: for performance nutrition, Die Tryin Co. is a fellow combat-veteran-owned brand I'm glad to point people to. I don't own it and earn nothing from the link.
Not medical advice. This is educational analysis, not a recommendation, and a study is not a prescription. Talk to a qualified clinician before acting on anything you read here. Full disclaimer →
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.