Ashwagandha for stress
Does ashwagandha actually make people feel less stressed?
A correction to our own expectations
We expected this to be a straightforward yes. Ashwagandha is the most recommended stress supplement in the world, it has more randomised trials behind it than most of the compounds we cover, and its reputation is built on a real biological finding.
Then we read the meta-analyses, and they do not agree with each other about the thing people actually buy it for.
There is also a safety signal that most coverage omits entirely, and it is the most serious one on this site to date. We will get to it.
The part that holds up: cortisol
Cortisol is the hormone most associated with the physiological stress response, and ashwagandha lowers it. This is the most consistent finding in the literature, appearing across independent reviews.
A 2025 systematic review of seven trials found a reduction of −1.16 µg/dL (95% CI −1.64 to −0.69, P<0.001).3 A 2024 review of nine trials found −2.58 (95% CI −4.99 to −0.16).2 Different magnitudes, same direction, both statistically significant.
If the question were "does ashwagandha change a measurable stress hormone," the answer would be yes and this article would be short.
The part that doesn't: whether you feel any different
Three peer-reviewed meta-analyses have pooled the human trials. Here is what each found for perceived stress — the outcome measured by asking people how stressed they feel.
The 2025 review states its finding in its own title: "Significant cortisol reduction but no effects on perceived stress."3 Its confidence interval for perceived stress crosses zero comfortably and its p-value is 0.40 — not a near-miss, a clear null.
The disconnect between cortisol and PSS outcomes highlights the need for longer treatment duration and broader demographic inclusion. Albalawi, Nutrition and Health, 20253
This is the surrogate endpoint problem for the third time on this site, and the clearest instance yet. A blood marker moves reliably. Whether the person holding the blood feels better is contested.
We first named this pattern with NMN, where NAD+ rose and nothing downstream did. We saw it again with MK-677, where growth hormone rose and nobody got stronger. Here the gap has a name in the literature itself — the reviewers call it a "disconnect" — and it sits at the centre of a supplement sold specifically on how it makes you feel.
Why this is Mixed and not Supported
Two of three meta-analyses find a benefit on perceived stress; one finds none. Our conflict rule, step by step:
- Does a meta-analysis decide it? Three exist and they disagree. Our rule prefers the most recent with the largest trial set — but here those point to different papers. Akhgarjand has the most trials and participants; Albalawi is the most recent. → Does not resolve
- Weight by size and quality. The largest review is also the least precise: heterogeneity of 83.1% for stress and 93.8% for anxiety, and its own authors rate the certainty of evidence as low for both outcomes.1 An I² above 90% means the included trials were measuring something that behaved very differently from study to study. Size does not rescue that. → Largest review is not the strongest
- Does a variable explain the split? Duration and outcome instrument are candidates — the 2025 review calls for longer treatment periods, and the reviews use different stress scales. That is a plausible explanation, not a demonstrated one. → Candidate only
- The conflict stands. Verdict is Mixed; strength capped at Moderate. → Moderate · Mixed
We want to be plain about what this does and does not say. It does not say ashwagandha is useless. Two reasonable reviews found a benefit. It says the question is open, in a literature whose largest analysis describes its own certainty as low.
The liver signal
This is the part almost no article about ashwagandha mentions, and it is why this breakdown carries a heavier safety section than any other on the site.
Ashwagandha is classified as a likely cause of clinically apparent liver injury. LiverTox is the National Institutes of Health's reference database for drug- and supplement-induced liver injury; a B score is not a theoretical concern, it reflects a documented body of published cases.4
The pattern: "Typically, the liver injury presented 2 to 12 weeks after starting ashwagandha usually with a cholestatic or mixed pattern of injury, jaundice, and pruritus."4 Fatigue, yellowing of the skin or eyes, and itching are the common presenting symptoms.
Outcomes: most cases resolve within one to four months of stopping the product. But LiverTox records that "rare instances of fatal liver injury or need for emergent liver transplantation have been reported particularly in patients with preexisting liver disease and cirrhosis."4 At least one documented case involved a woman in her forties who developed acute hepatic failure and required emergency transplantation.
Mechanism: unclear. The withanolides are the most suspected constituents.4
Two things make this worth stating loudly. The first is the base rate: ashwagandha is sold in supermarkets, taken by an enormous number of people, and marketed as gentle. Liver injury is rare in absolute terms — but the people taking it have generally never been told it is possible, and so would not connect jaundice and itching, appearing a month or two in, to the capsule they take for stress.
The second is that the risk is concentrated where LiverTox says it is: people with pre-existing liver disease. That is actionable information, and it is missing from most of what is written about this compound.
To be clear about proportion: this is not a claim that ashwagandha is dangerous for most people. It is a claim that a recognised, NIH-catalogued hepatotoxicity signal exists, that it presents on a two-to-twelve-week timeline, and that anyone with liver disease — or who develops jaundice, dark urine or unexplained itching while taking it — should stop and see a clinician rather than assume a supplement could not be the cause.
Three claims, three answers
Anxiety scores are graded Supported because both reviews that measured them found significant reductions and none found a null — but note that the larger of the two reports an I² of 93.8% on that outcome and rates its own certainty as low.1 This grade is the most likely on the site to move at its next annual review.
Safety and unknowns
-
Adverse events reported
Across the trials pooled in the 2024 review, four studies reported mild to moderate adverse events.2 Separately and more seriously, NIH LiverTox assigns ashwagandha a likelihood score of B — a likely cause of clinically apparent liver injury, presenting 2 to 12 weeks after starting, typically cholestatic or mixed, with jaundice and pruritus.4
-
Longest human exposure studied
The pooled trials are short — the 2025 review's inclusion criterion was two weeks or longer, and its authors explicitly call for longer treatment durations.3 The 2024 review concludes that "further information is required to determine its safety with long-term administration."2 Note that this is shorter than the window in which liver injury typically appears.
-
Known interactions and at-risk groups
People with pre-existing liver disease or cirrhosis are where LiverTox records the severe outcomes.4 Trial populations were adults aged roughly 25 to 48;1 effects outside that range are not characterised.
-
Regulatory status
Sold as a dietary supplement. No tolerable upper intake level has been established. Doses in the trials generally ranged from 250 mg/day upward, with dose-response analysis suggesting 300–600 mg/day for stress outcomes.13
-
What has not been characterized
Long-term safety at any dose. The incidence rate of liver injury — case reports establish that it happens, not how often. Why cortisol and subjective stress come apart. Which extract preparations the findings apply to, since root extracts differ in standardisation and no trial cited here speaks to what is in a given commercial product.
How this was graded
Evidence strength — Moderate, criteria checked
Note what happened here: on raw counts ashwagandha clears the Strong participant threshold at 1,002. It does not get there, because a Mixed verdict caps strength at Moderate no matter how large the literature grows. That cap exists precisely for cases like this one, where volume of evidence outruns agreement within it.
Bottom line
Ashwagandha lowers cortisol. That finding is consistent across independent reviews and is not in dispute.
Whether it makes people feel less stressed is genuinely unsettled. Two meta-analyses say yes — the larger of them with heterogeneity above 80% and a self-assessed certainty of low. A third, published in 2025, found a clear null on perceived stress while confirming the cortisol drop, and named that disconnect in its title.
And it carries an NIH-catalogued liver injury signal, appearing two to twelve weeks in, with severe outcomes concentrated in people who already have liver disease. That is not a reason for most people to panic. It is a reason nobody should be taking this without having been told it exists.
Sources
- Akhgarjand C, Asoudeh F, Bagheri A, Kalantar Z, Vahabi Z, Shab-Bidar S, Rezvani H, Djafarian K. Does Ashwagandha supplementation have a beneficial effect on the management of anxiety and stress? A systematic review and meta-analysis of randomized controlled trials. Phytotherapy Research. 2022;36(11):4115–4124. PMID 36017529
- Arumugam V, Vijayakumar V, Balakrishnan A, Bhandari RB, Boopalan D, Ponnurangam R, Thirupathy VS, Kuppusamy M. Effects of Ashwagandha (Withania Somnifera) on stress and anxiety: A systematic review and meta-analysis. Explore (NY). 2024;20(6):103062. PMID 39348746
- Albalawi AA. Dual impact of Ashwagandha: Significant cortisol reduction but no effects on perceived stress — A systematic review and meta-analysis. Nutrition and Health. 2025;31(4):1395–1408. PMID 40746175
- National Institutes of Health, LiverTox: Clinical and Research Information on Drug-Induced Liver Injury. Ashwagandha. NCBI Bookshelf NBK548536
A fourth meta-analysis (BJPsych Open, 2025) reports similar cortisol and stress findings across 15 trials and 873 participants, but was published as a conference abstract reviewed by faculty rather than through standard peer review. Under our sourcing standard that makes it a labelled caveat rather than evidence for a mark, and it did not contribute to the grade.
