Ashwagandha, Cortisol, Sleep and Anxiety, Trial by Trial

In the published human trials, groups taking a standardised ashwagandha extract finished with lower morning cortisol readings than the groups taking placebo. Chandrasekhar and colleagues (2012) recorded a 27.9 per cent fall in serum cortisol over 60 days against 7.9 per cent on placebo, in 64 adults with a history of chronic stress. That single result is what almost every page about ashwagandha and cortisol is quoting, usually without saying so.

What those pages leave out is the measurement. Every cortisol figure in this literature comes from one blood sample taken in the morning, in a trial lasting 60 days or less, in fewer than 140 people, and in most cases paid for by the company that owns the extract. This page gives you each trial with its dose, its population, its exact cortisol measurement and its funder, then the sleep and anxiety scores, then what nobody has measured yet.

What cortisol is, and what a normal day looks like

Cortisol is a steroid hormone made in the cortex of the adrenal glands. The Society for Endocrinology describes its control as a loop: the hypothalamus releases corticotrophin-releasing hormone, the pituitary responds with adrenocorticotropic hormone, the adrenal cortex responds with cortisol, and rising cortisol then blocks further release of corticotrophin-releasing hormone. That loop is the hypothalamic-pituitary-adrenal axis, usually written as the HPA axis. It is the thing the word adaptogen is gesturing at, and it is described more fully in our guide to what an adaptogen is.

The second fact matters more for reading a study. Cortisol is not a level. It is a curve. It is highest in the morning shortly after waking and falls across the day to a low point around midnight, and in people who work night shifts the curve moves with the shift rather than with the clock.

Two consequences follow, and they decide how much any number is worth.

A single blood sample is one point on that curve. Whether it looks high or low depends on what time you woke, how long you had been awake, and whether the tube was filled at 07:30 or at 09:15. Move the draw an hour and you move the number, with nothing having changed in the person.

And a lower cortisol reading is not automatically a good result. Cortisol that is too low is a serious medical problem in its own right, with fatigue, dizziness on standing and muscle weakness among the recognised signs. There is no target figure a supplement should be moving you towards, and any page that implies one is selling you something.

How we read a cortisol trial

Every trial below is described against the same six criteria, because these are the six things that decide whether a cortisol number means anything at all.

The sample. Serum from a venous draw, or saliva. They measure different things: salivary cortisol reflects the unbound fraction and is not affected by carrier-protein levels, so the two are not interchangeable numbers.

The timing. What time of day, how long after waking, fasting or not. In a hormone with a daily curve, timing is not a detail. It is half the measurement.

How many samples. One morning draw gives you a point. A day of repeated samples gives you a slope. Only the second one can tell you whether the shape of somebody’s day changed.

The population. Every trial here recruited healthy adults reporting stress, with psychiatric diagnoses, chronic illness and medication excluded. That exclusion list is why none of these results transfers to a diagnosed condition.

The extract. KSM-66, Sensoril and Shoden are different preparations at different withanolide specifications, and the trials are not testing the same material. Which extract is on your label, and what the milligram figure refers to, is covered in ashwagandha dosage and timing.

Who paid. Most of these trials were funded by the owner of the extract tested, or used product supplied free by them. That does not make a result wrong. It does mean the field has almost no independently funded replication, and you should know which is which.

What each ashwagandha cortisol trial actually measured

Four trials carry essentially all of the cortisol evidence. Here they are in order, with the numbers as published.

Chandrasekhar, Kapoor and Anishetty (2012), in the Indian Journal of Psychological Medicine, randomised 64 adults aged 18 to 54 with a history of chronic stress to 300mg of a full-spectrum root extract twice daily or placebo for 60 days. Sixty-one completed. Serum cortisol was drawn in the morning on day 0 and day 60. The extract arm fell 27.9 per cent from baseline against 7.9 per cent on placebo, at P equals 0.002. Perceived Stress Scale, Depression Anxiety Stress Scale and General Health Questionnaire scores were also lower. The paper declares no source of support and no conflict of interest, which is unusual in this literature and worth noting. The authors state plainly that the sample was not very large and ask for longer studies in more diverse populations.

Salve, Pate, Debnath and Langade (2019), in Cureus, ran three arms in 60 healthy adults aged 18 to 55 who scored 20 or more on the Perceived Stress Scale: KSM-66 at 250mg a day, KSM-66 at 600mg a day, and placebo, for 8 weeks. Fifty-eight completed. Morning serum cortisol was taken at baseline, week 4 and week 8. On 600mg it went from 16.12 to 10.86 micrograms per decilitre; on 250mg from 16.30 to 13.61; on placebo from 16.15 to 15.52. Perceived stress scores on 600mg went from 22.95 to 14.15, against 22.70 to 16.63 on placebo. A seven-point self-rated sleep score ended at 3.05 on 600mg against 4.89 on placebo, where a lower number is better sleep. The extract was manufactured and gifted by Ixoreal Biomed.

Lopresti, Smith, Malvi and Kodgule (2019), in Medicine, gave 60 adults aged 18 to 65 with mild anxiety, defined as a Hamilton Anxiety Rating Scale score of 6 to 17, either 240mg of Shoden daily or placebo for 60 days. Morning serum cortisol fell 23 per cent in the extract arm against a 0.5 per cent rise on placebo. DHEA-S, an adrenal androgen precursor, fell 8.2 per cent against a 2.5 per cent rise, which is a reminder that the axis does not move one hormone in isolation. Hamilton scores fell 41 per cent against 24 per cent. The project was funded by Arjuna Natural Ltd, which supplied the extract.

Pandit and colleagues (2024), in Nutrients, is the largest and the most informative. One hundred and thirty one chronically stressed adults, mean age 35, were randomised to Sensoril at 125mg, 250mg or 500mg or to placebo, taken as a single capsule at night for 8 weeks; 98 completed per protocol. At week 8 plasma cortisol was 12.34 micrograms per decilitre on placebo, 8.61 on 125mg, 9.65 on 250mg and 7.90 on 500mg. This trial also measured adrenocorticotropic hormone, the pituitary signal one step upstream, which ended at 16.52 picograms per millilitre on placebo against 7.37, 7.77 and 8.55 on the three doses. It is the only trial here that looked above the adrenal gland. It was funded by Natreon, the extract’s owner, and lost a large number of participants to the pandemic.

Trial	Extract and daily amount	Duration	Cortisol measure	Funding
Chandrasekhar 2012, 64 adults	KSM-66, 600mg split	60 days	Morning serum, day 0 and day 60	None declared
Salve 2019, 60 adults	KSM-66, 250mg or 600mg	8 weeks	Morning serum, three timepoints	Extract gifted by Ixoreal
Lopresti 2019, 60 adults	Shoden, 240mg	60 days	Morning serum	Arjuna Natural Ltd
Pandit 2024, 131 adults	Sensoril, 125, 250 or 500mg	8 weeks	Morning plasma, plus ACTH	Natreon Inc, the extract owner

Pooled, the picture is consistent and small. Bachour and colleagues (2025), in BJPsych Open, combined 15 randomised trials covering 873 participants and reported a mean difference in cortisol of about 2.36 units at 8 weeks, with a 95 per cent confidence interval running from 3.26 to 1.46, alongside lower perceived stress and lower anxiety scores. Consistency across small manufacturer-funded trials of a handful of proprietary extracts is a weaker signal than it looks, and the honest reading is a real but modest difference measured in one narrow way.

What to take from this section: the number you have read elsewhere is a morning blood draw in fewer than 140 people over 8 weeks. It is not nothing, and it is not a description of your day.

The measurement gap nobody mentions

Here is the thing that separates a cortisol reading from a cortisol curve, and it is missing from every page currently ranking for this question.

Endocrinology has three standard ways of describing the daily rhythm: the cortisol awakening response, which is the rise across the first 30 to 45 minutes after waking; the diurnal slope, the rate at which cortisol falls across the day; and the total area under the daytime curve. All three need repeated saliva samples across one or more full days.

Ryan and colleagues (2016), in Annals of Behavioral Medicine, reviewed 78 randomised trials that used salivary diurnal cortisol as an outcome. The median trial collected saliva on a single day, 73.1 per cent of them on one day only. The median number of samples was four a day, ranging from two to nine. Only 62.8 per cent included a sample at the moment of waking, which the awakening response cannot be calculated without. Just half measured any rhythm parameter, and 2.6 per cent measured all three. Their conclusion is direct: salivary diurnal cortisol is measured inconsistently across trials, which limits the interpretation of findings within and between studies.

Now put the two facts side by side. The ashwagandha trials did not use those methods at all. They used one morning venous sample. So the sentence the evidence supports is that a morning cortisol reading was lower after 8 weeks in these groups than in the placebo groups. The sentence it does not support is that anyone’s daily cortisol rhythm changed shape, because that was never measured.

This is also why a home cortisol kit is unlikely to tell you much. A single spot reading, taken at whatever hour was convenient, is the weakest version of an already noisy measurement. If the question matters to you, ask a doctor who can control the timing and read the result against the rest of your picture.

Sleep, measured rather than promised

Sleep is the outcome with the most direct evidence, because it was measured with questionnaires and clock times rather than inferred from a hormone.

Langade and colleagues (2019), in Cureus, randomised 60 adults aged 18 to 60 with diagnosed insomnia in a two to one ratio to 300mg of KSM-66 twice daily or placebo for 10 weeks. At week 10, sleep onset latency was 29.00 minutes in the extract arm against 33.95 minutes on placebo, sleep efficiency 83.49 per cent against 79.68 per cent, and Pittsburgh Sleep Quality Index scores 9.15 against 11.84. Read the first of those carefully: it is a difference of about five minutes in average time to fall asleep. No cortisol was measured in this trial at all, which is worth knowing given how often sleep and cortisol are quoted in the same breath.

Cheah and colleagues (2021), in PLOS ONE, pooled five trials with 400 participants and found a standardised mean difference for overall sleep of 0.59 in favour of ashwagandha, with a 95 per cent confidence interval from 0.75 to 0.42 and heterogeneity of 62 per cent. The subgroups all pointed the same way: larger differences at 600mg a day or more than below it, larger at 8 weeks or more than at shorter durations, and larger in people with diagnosed insomnia than in healthy adults. The authors note that data on serious adverse effects are limited and that long-term use is unassessed.

So the answer to whether ashwagandha helps you sleep is that in these trials the extract groups recorded modestly better sleep measures than the placebo groups, more so in people who already slept badly, over 8 weeks or longer. Anyone describing a sedative effect from the first night is describing something no trial has measured.

Anxiety scales, and what they are not

People search for ashwagandha for stress and anxiety, and for whether it makes them calm, far more often than they search for the hormone. Every anxiety figure in this literature is a score on a rating scale in people who were screened to be free of psychiatric illness. That constraint is what makes the numbers honest and what makes them limited.

Lopresti’s 2019 trial recruited only people with Hamilton scores of 6 to 17, which is mild territory by design, and reported a 41 per cent fall against 24 per cent on placebo. The Depression Anxiety Stress Scale short form fell 30 per cent against 10 per cent, a difference that did not reach statistical significance and is therefore not a result. Langade’s insomnia trial recorded Hamilton scores of 18.49 against 21.53 at week 10. The Bachour pooled analysis found lower Hamilton scores at 8 weeks across the trials it combined.

None of that is evidence about a diagnosed anxiety disorder, because people with one were excluded from every trial. If you are being treated for anxiety, or taking a sedative or an antidepressant, the question to ask is an interaction question and it belongs with your doctor or pharmacist, not with a shop.

The evening stack, and what has never been tested together

Ashwagandha is often stacked with other things in an evening routine, and it is worth being exact about what each part has behind it.

Ashwagandha has the cortisol and sleep trials above. Reishi has its own small human literature and is widely sold as a calming mushroom despite not being a sedative, which we set out in our reishi page. Shilajit has about eight human trials, none of which measured cortisol, sleep or anxiety at all, as our audit of the shilajit trials shows.

That last point is the one to hold onto. Our capsules pair 250mg of a named active with 250mg of shilajit extract, and there is no trial of that combination, or of any ashwagandha plus shilajit combination, on any of the outcomes on this page. The reason to pair them is that you want both materials in one capsule rather than two jars, and the honest description of the pairing is on our ashwagandha and shilajit page. A combination is not entitled to inherit the evidence of its parts.

What is not established, and who should not take it

This is the part the marketing pages skip, and it is the most useful section here.

Nothing above establishes an effect in anyone who was not in those trials. No trial ran longer than 10 weeks on these outcomes, so nothing is known about a year of continuous use. No trial measured the daily cortisol curve. No trial enrolled anyone with a diagnosed anxiety disorder or depression, and only one enrolled people with diagnosed insomnia.

Ashwagandha also has a genuine caution list, and each item below is set out properly on our ashwagandha benefits and side effects page rather than repeated here. Liver injury case reports exist and are the reason several European authorities have assessed the ingredient. Denmark’s position is the strictest in Europe, and that page gives the dates and the named assessments. Thyroid medication is a real interaction, because trial data show the extract moves thyroid indices. Sedatives and medicines acting on the central nervous system are an interaction question by definition when a substance is being taken at night. Autoimmune conditions appear on the exclusion lists of the European assessments. What the trials recorded about appetite and body weight sits on our ashwagandha and body weight page.

Do not take it in pregnancy or while breastfeeding. Do not give it to children. If you take prescribed medication of any kind, or you have a diagnosed condition, ask a doctor or pharmacist before you start, and stop if anything changes.

What actually decides this

Three things decide whether ashwagandha is worth your money for this purpose, and none of them is the 27.9 per cent figure. The first is whether a single morning blood measurement in 64 people is the kind of evidence you find persuasive. The second is whether the outcome you actually care about, which is usually sleep or how a stressful week feels, was measured directly rather than inferred from a hormone. The third is whether you will run it for the 8 weeks the trials ran, because there is no evidence at all about week one.

If that reads as an argument for patience rather than an argument for a purchase, it is meant to. When you do want the material, our ashwagandha and shilajit capsules hold 250mg of ashwagandha alongside 250mg of shilajit extract, and the amount per capsule is on the label so you can compare it against the trial figures above yourself. Read the dosage page first, pick your eight weeks, and write down the date you start.