Does ashwagandha work? What the trials measured, and what they did not

Ashwagandha is the most searched plant on this shelf, and the search is nearly always the same three words: does it work. That expects a yes or a no. What follows is neither, because the question as written cannot produce one - and the reason is more useful than either answer.

Three things are tangled in those three words: what has been measured in trials, what a trial of sixty people can establish, and what a regulator has decided. Only the third governs what a British pack may say to you.

What do the ashwagandha studies actually measure?

Start with what has actually been done, because it is more specific and less dramatic than the internet suggests. Almost all the human research quoted in supplement marketing sits in three areas - perceived stress, anxiety and sleep - and nearly all of it is measured with a rating scale rather than a machine.

The scales are worth knowing by name: they are the actual finding in these studies. The Perceived Stress Scale is ten questions about the previous month, scored 0 to 40. The Hamilton Anxiety Rating Scale is fourteen items scored by a clinician, 0 to 56. The Pittsburgh Sleep Quality Index is a self-reported sleep questionnaire. Sleep onset latency - how long it takes to fall asleep - is sometimes taken by a wrist device, the closest any of this gets to an objective instrument. When a headline says a trial 'found an effect', it means a number on one of those questionnaires moved further in one group than the other.

Three trials come up repeatedly. A 2012 study in the Indian Journal of Psychological Medicine randomised 64 adults at one hospital in Hyderabad, gave 300 mg of concentrated root extract twice daily for 60 days, and reported Perceived Stress Scale scores falling 44.0 per cent against 5.5 per cent on placebo. A 2019 trial in Cureus took 60 stressed but healthy adults for eight weeks at 250 mg or 600 mg a day, and reported lower Perceived Stress Scale scores at both doses but a difference on the Hamilton scale only at 600 mg. A second 2019 Cureus trial followed 60 people with insomnia for ten weeks on 300 mg twice daily, and reported Pittsburgh index scores moving from 13.08 to 9.15 against 13.47 to 11.84 on placebo.

What a sixty-person trial can and cannot settle

Four features of that list matter more than the percentages do. The samples are 60 to 64 people, the size at which one unusual participant shifts a group average. The trials ran eight to ten weeks, so nothing in them speaks to a year. They were run at single centres, mostly in India, in volunteers screened free of other illness. And in each case the extract came from the company that sells it - normal in this field, disclosed in the papers, still worth holding in view.

One more feature does the most damage to the search query. Every trial above tested a concentrated root extract standardised to a stated percentage of marker compounds, at a specified daily amount. None of them tested ashwagandha. There is no such thing as ashwagandha in the sense the question assumes: there is root powder, root extract and root-and-leaf extract, traded at withanolide specifications from around 2.5 per cent to 35 per cent. Two packs both saying ashwagandha on the front can be different materials by a factor of ten.

Why isn't a positive trial the same as approval?

A trial and a regulatory evaluation are not the same exercise at different scales, and the difference explains most of the argument around this plant. A trial asks a narrow question: in this group, on this preparation, at this dose, for this long, did a number on this scale move further than in the placebo group, by more than chance comfortably explains? That is a real question, and a well-run trial answers it honestly - about that preparation, that dose, that group, that scale and that duration.

A regulatory assessment asks something wider: whether the whole body of evidence establishes a cause-and-effect relationship between a sufficiently characterised substance and a defined beneficial physiological effect, in the general population, at an amount a person could reasonably consume. 'Sufficiently characterised' is where most botanical dossiers stall, because an extract has to be pinned to a specification before anyone can rule on it. 'The general population' rules out a finding that only holds in screened volunteers at one hospital. And the outcome is a fixed form of words, tied to fixed conditions of use, which any company meeting them may print and none may strengthen.

Why 'there are studies' is not 'it is approved'

The gap between the two is not one being stricter about the same thing. A trial can be immaculate and still leave the regulatory question untouched, because it was never designed to reach it. Which is why 'there are studies' and 'it is approved' are statements about different objects, and why a brand quoting the first while implying the second is doing something dishonest with a true sentence.

What do the systematic reviews of ashwagandha say?

When individual trials are small, the standard move is to pool them. A 2021 meta-analysis in PLoS ONE gathered five randomised trials covering 400 participants, at 120 mg to 600 mg a day over six to twelve weeks, and reported a pooled standardised mean difference on overall sleep of -0.59, 95 per cent confidence interval -0.75 to -0.42. The authors called that small but statistically significant, recorded substantial disagreement between the trials, noted that only two had been registered in advance and one was partly sponsored by a manufacturer, and stated plainly that data on serious adverse effects were limited. A 2024 review in Explore pooled nine trials and 558 patients, reporting a mean difference of -4.72 points on the Perceived Stress Scale, interval -8.45 to -0.99, and -2.19 on the Hamilton scale, interval -3.83 to -0.55. A 2024 review in Human Psychopharmacology pooled five trials and 254 participants and reported a mean Hamilton difference of -5.96, interval -10.34 to -1.59, with a heterogeneity statistic of 98 per cent, concluding that larger studies are needed before the effects on anxiety and insomnia can be confirmed.

Confidence intervals and 98 per cent heterogeneity, in plain English

Two of those numbers deserve translating. A confidence interval is the range of values the data are compatible with, so a pooled stress finding running from -8.45 to -0.99 on a forty-point scale is compatible both with a difference large enough to notice and with one small enough to be invisible in a life. The heterogeneity statistic measures how far the trials disagree with each other beyond what chance would produce. At 98 per cent, they are not really measuring the same thing in the same way, and the average of them is an average of unlike objects.

The American National Center for Complementary and Integrative Health, which has no product to sell, puts the position in three sentences worth reading as written. Many clinical trials have looked at ashwagandha, but many of the studies have had small sample sizes and have used a variety of ashwagandha preparations. Some ashwagandha preparations may be effective for insomnia and stress. Evidence is unclear about its effects on anxiety.

Read the middle sentence again and notice the qualifier. Not ashwagandha - some ashwagandha preparations. The same body records that use of up to about three months appears tolerated, that long-term safety is not established, and that cases of liver injury have been linked to ashwagandha supplements, rare but documented. That belongs in the answer too, because 'does it work' and 'is it fine' are usually one thought.

Why is there no authorised ashwagandha health claim in the UK?

Now the part that decides what a British label may print, and it has almost nothing to do with the trials above. Between July 2008 and March 2010, some 4,637 general function health claims went to the European Food Safety Authority. By June 2011 it had published 341 opinions covering 2,758 of them. The remainder - 1,548 claims relating to botanicals - were put on hold by the European Commission and member states pending a final decision.

'On hold' since 2010: neither authorised nor rejected

The queue has not moved since. Ashwagandha's claims sit in it, with those for every other plant in our pack. They were not examined and found wanting. They were never examined. Great Britain inherited the position and now keeps its own register, and the rule is simple: only a claim listed as authorised there may be used, and no claim for ashwagandha is listed.

So a British pack may tell you the species, the plant part, the form and the milligrams, and may describe traditional use under a footnote saying the wording is not authorised. It may not tell you what the plant does to your body, whatever the trials say, because the permission to make that statement is a document, and the document does not exist. Herbal supplements here are regulated as food rather than licensed as medicines, and food law is stricter about promises than most people assume.

So the absence of an authorised claim is not a verdict against the plant, and it is not a loophole for the brand. It is the state of a piece of paperwork in a process that stalled fifteen years ago. Read as proof the plant does nothing, it is overread. Read as room to promise something, it is misread the other way.

Which brings this close to home, so here is our number. The trials above used 240 mg to 600 mg a day of concentrated root extract. Our tablet declares Ashwagandha 10 mg, one of six botanicals adding to 100 mg, with a maximum of one tablet in any 24 hours. That is a small fraction of the amounts in those studies, and not the same kind of material either - a whole-weight botanical rather than a standardised extract, which makes the two figures hard to compare at all rather than merely far apart.

We print the 10 mg to the milligram for exactly this reason. A pack that hides its ashwagandha inside a proprietary blend total has made that comparison impossible to run. Ours makes it easy. That is not a defence of the number and we will not dress it as one - it is the number, published so you can do the arithmetic yourself.

If you came here for a verdict, decide first what would count as one. Trial evidence on named extracts at named doses exists and is described above, sample sizes attached. A regulatory finding does not exist, here or in the European Union, and no pack can supply it. What a pack can supply is the species and the plant part, the form and the weight, a batch code and a certificate of analysis - and the plant is a nightshade sold under a borrowed name, so the species line matters here. Quote the batch code and best-before date from the blister foil to hello@maleup.co.uk and we will send the paperwork for that run.

Does ashwagandha work is a question about a document as much as a plant. The plant has been studied a fair amount and settled very little. The document was never written. The rest of the shelf is people filling that silence with a sentence they are not entitled to.

Common questions

† Describes the traditional use of these botanicals (UK 'on-hold' botanical claims under retained Regulation (EC) 1924/2006). Not authorised EFSA health claims. AllMaleUp is a food supplement, not a medicine.