If you have any liver condition — current or past — do not take ashwagandha. The TGA has issued two safety advisories about rare liver injury, most recently in July 2026. The published case series are mostly recoverable, but every death occurred in someone who already had chronic liver disease. Warning signs: yellowing of the skin or eyes, dark urine, itching, nausea, unusual fatigue, pain under the right ribs.
The short version: "600 mg of ashwagandha" on a label tells you almost nothing. The extracts differ in potency by more than twentyfold, and independent testing suggests commercial withanolide content can deviate from label claim by two to thirty-five times. Meanwhile the best-designed stress trial ever run on the herb failed its primary endpoint, and the 2026 meta-analysis of resistance-training trials concluded the evidence "does not establish that adjunctive ashwagandha meaningfully enhances" anything.
There is good news for Australian buyers, though: the two products here that do deliver a trial-equivalent dose are also among the cheapest, at 65 to 75 cents a day.
The extracts are not interchangeable
| Plant part | Standardisation | Typical dose | |
|---|---|---|---|
| KSM-66 | Root only, aqueous extraction | ≥5% withanolides | 300–600 mg/day |
| Sensoril | Root and leaf | ≥10% withanolides | 125–250 mg/day |
| Shoden | Root and leaf | ≥35% withanolide glycosides | 120–240 mg |
| Witholytin | Root | 1.5% total withanolides | 400 mg/day |
| Unstandardised root powder | Root (usually) | None | 1–5 g |
Why the doses differ fivefold: it's withanolide delivery, not extract mass. 600 mg of KSM-66 at 5% is about 30 mg of withanolides. 250 mg of Sensoril at 10% is about 25 mg. They converge.
(Note that Shoden's ≥35% refers to withanolide glycosides — a different analyte from "total withanolides," so the numbers aren't directly comparable.)
And here's why "standardised" alone isn't enough. Witholytin is a named, standardised, clinically trialled extract — at 1.5% withanolides. At 400 mg/day that's about 6 mg of withanolides, roughly one-fifth of the KSM-66 trial dose. Standardisation tells you the number is measured. You still need the percentage.
On withaferin A and the root-versus-leaf argument
Withaferin A is the most cytotoxic withanolide, and it's concentrated in leaf rather than root. That's the basis of the entire regulatory fight — India's AYUSH ministry mandated root-only in April 2026, and FSSAI banned leaf in all food forms.
But be careful with this argument, because it is not what the liver evidence shows. The mechanistic hypothesis in the literature points at withanone, not withaferin A, and the Dutch RIVM assessment notes that withanone concentration in commercial extracts is simply unknown.
More to the point: no case series links withaferin A content, or leaf content, to any observed human liver injury. Several of the implicated products were KSM-66 — root-only, aqueous, the "safe" one. The leaf story is a plausible mechanism and a regulatory argument. It is not a demonstrated cause, and writing it as one would be the biggest available error on this topic.
What the trials actually show
Stress and cortisol
**Chandrasekhar K, Kapoor J, Anishetty S. Indian J Psychol Med 2012;34(3):255–262** is the trial behind most ashwagandha marketing. n=64 randomised, 61 completed, KSM-66 300 mg twice daily, 60 days.
- Perceived Stress Scale −44.0% vs −5.5% placebo, p<0.0001
- Serum cortisol −27.9% vs −7.9%, p=0.002
- DASS depression subscale −77.0% vs −5.2%
Two things to flag. The paper declares "Source of Support: Nil" while testing a proprietary branded extract supplied by its manufacturer — a declaration that is formally present and substantively incomplete. And a 77% reduction on a depression subscale in a non-depressed population is not a credible effect size.
The better trial failed. Smith SJ, Lopresti AL, Fairchild TJ. J Psychopharmacol 2023;37(11). n=120, adults 40–75 with high stress and fatigue, Witholytin 200 mg twice daily, 12 weeks — the best-powered and best-reported stress trial in the literature.
Primary outcome, Perceived Stress Scale: p=0.867. Null.
Fatigue improved (p=0.016) and heart rate variability improved (p=0.003), but the thing it was designed to measure didn't move. The trial was funded by the manufacturer, which makes the null result more credible, not less.
A 2026 dose-response meta-analysis of 22 RCTs (n=1,391) rated overall certainty GRADE "very low." (We're not quoting its effect sizes: the reported standardised mean differences of −5.88 and −6.87 are arithmetically impossible — an SMD of nearly 7 means seven pooled standard deviations — and indicate a mis-specified model.) One finding from it is worth keeping: lower doses (≤500 mg/day) showed larger effects than higher ones, which contradicts the commercial "more is better" framing.
Sleep — the most honest number in the whole literature
**Cheah KL, Norhayati MN, Husniati Yaacob L, Abdul Rahman R. PLOS ONE 2021;16(9):e0257843. 5 RCTs, n=400. All five conducted in India.**
- Sleep quality SMD −0.59 (95% CI −0.75 to −0.42) — which the authors themselves call "small"
- Insomnia subgroup SMD −0.84
- Quality of life: no significant improvement in any domain
**Langade D, Kanchi S, Salve J, et al. Cureus 2019;11(9):e5797** gives the number people should actually see. n=58, primary insomnia, KSM-66 300 mg twice daily, 10 weeks:
Sleep onset latency: 29.0 minutes versus 33.95 minutes on placebo (p=0.019).
That's a statistically significant improvement of about five minutes, against marketing that implies transformation. Sleep efficiency and PSQI also improved. (Note the unbalanced groups — 39 versus 19 — and the same "no financial support received" declaration alongside manufacturer-supplied extract.)
Testosterone — read the population, not the headline
No trial demonstrates a clinically meaningful testosterone increase in a healthy, young, eugonadal man. Every positive trial is in a selected population.
**Lopresti AL, Drummond PD, Smith SJ. Am J Mens Health 2019;13(2).** (Commonly miscited as a KSM-66 trial — it used Shoden.) n=57 enrolled, 43 completed. Population: overweight men aged 40–70 with self-reported fatigue, explicitly not healthy young men.
- Testosterone +14.7%, p=0.010 — 309.99 → 355.57 pmol/L
- Cortisol −7.8%, p=0.319 — null
- Funded by the manufacturer
Read the clinical framing: 310 → 356 pmol/L is within the normal reference range at both ends. A 14.7% move inside the reference range is not an intervention.
**Wankhede S, Langade D, Joshi K, et al. JISSN 2015;12:43 is the trial every testosterone booster cites. n=50 healthy, resistance-training-naïve** men, KSM-66 300 mg twice daily, 8 weeks:
- Testosterone +96.2 ng/dL vs +18.0 ng/dL placebo, p=0.004
- Bench press 1RM +46.0 kg vs +26.4 kg, p=0.001
A 46 kg bench press gain in eight weeks is at the extreme edge of plausibility even allowing for novice gains — and the placebo arm's +26.4 kg is itself very high. Treat the measurement credibility as the issue, not just the funding.
Strength and body composition — the best-quality assessment available
***Nutrients 2026;18(17):2815 — systematic review and random-effects meta-analysis of adjunctive ashwagandha with resistance training. Only 3 RCTs, n=161.*
- Bench press 1RM SMD 0.60 (0.20–1.01), p=0.004
- Lower-body 1RM SMD 0.52 (0.19–0.86), p=0.002
- Body fat %: not significant
- Under small-sample adjustment, both strength estimates lost significance
- GRADE: Low for strength
- "All three included trials received funding or product supply from ashwagandha manufacturers" — the meta-analysts say this themselves
Their conclusion, worth quoting directly: "Current evidence does not establish that adjunctive ashwagandha meaningfully enhances resistance-training adaptations."
(A 2021 Bayesian meta-analysis reported a cardiorespiratory fitness effect of SMD 1.89. An SMD of 1.89 for VO2max is not a believable supplement effect; it came from small trials and required outlier removal. Treat it as evidence of a literature problem, not of aerobic benefit.)
Thyroid — a real, measured effect with a real safety implication
**Sharma AK, Basu I, Singh S. J Altern Complement Med 2018;24(3):243–248. n=50 with subclinical hypothyroidism**, 300 mg root extract twice daily, 8 weeks:
- T3 +41.5% at week 8 (p<0.0001)
- T4 +19.6% at week 8 (p<0.0001)
- TSH significantly decreased versus placebo (p=0.0002)
This is framed by the industry as a benefit, and in subclinical hypothyroidism it might be. But read it the other way: a 41.5% rise in T3 is dose-dependent thyroid stimulation. The RIVM assessment confirms consistent thyroid-hormone elevation in animals plus dose-dependent human T3/T4 increases, and the NIH Office of Dietary Supplements notes reports of thyrotoxicosis.
If you have Graves' disease or hyperthyroidism, or you take levothyroxine, carbimazole or thiamazole, ashwagandha is an unmonitored thyroid-active drug.
The funding pattern, reported honestly
Chandrasekhar 2012, Wankhede 2015 and Langade 2019 are all KSM-66 trials, all conducted in India, and all declare no funding while the manufacturer supplied the extract.
Lopresti 2019, Smith 2023 and the Sensoril trials are openly manufacturer-funded — and are also the better-reported and more conservative ones. Lopresti's cortisol result was null. Smith's primary outcome was null.
The honest reading is counterintuitive: the openly funded Western trials are more conservative and more completely reported than the "unfunded" ones. Disclosure isn't the problem. The undisclosed supply relationship alongside implausible effect sizes is.
The liver question
What the TGA has actually done
Two safety advisories. No mandated liver warning statement.
(If you've read elsewhere that Australia requires a liver warning on ashwagandha labels — we checked, and we could find no such mandate. There is a pre-existing mandatory pregnancy warning on certain preparations.)
February 2024 — first advisory. Covered two things: acute severe vomiting and diarrhoea that can be mistaken for infectious gastroenteritis, and rare liver injury. Advised that people with current or past liver disease avoid these products.
7 July 2026 — second advisory, and the numbers:
- 11 total Australian reports of possible ashwagandha-associated liver injury
- 4 new since the 2024 advisory; all four resolved on stopping, none hospitalised
- Of the 11: 6 had no other ingredient plausibly contributing; 4 of those 6 were ARTG-listed medicines; 2 of those 4 required hospitalisation
- Internationally, the TGA cites India — 8 cases, 3 deaths with single-ingredient preparations, and the Netherlands — 12 cases, onset from 6 weeks to 22 months
- As at 17 July 2026, 91 ARTG-listed complementary medicines contained ashwagandha, predominantly root extracts
- The TGA is actively evaluating evidence on different plant parts — the Australian position is live, not settled
Worth knowing as precedent: in September 2026 the TGA **removed Andrographis paniculata from the listed-medicines framework entirely** over allergy risk. The regulator will delist a popular herb when it decides to.
The published cases
**Björnsson HK, Björnsson ES, Avula B, et al. Liver International 2020;40(4):825–829.** Five cases — three from Iceland, two from the US Drug-Induced Liver Injury Network. Mean age 43.
- All five developed jaundice, with nausea, lethargy, itching, abdominal discomfort
- Latency 2–12 weeks
- Cholestatic or mixed pattern
- Itching and raised bilirubin were prolonged: 5–20 weeks
- No hepatic failure. Liver tests normalised in one to five months in four patients
- Chemical analysis confirmed ashwagandha; no other toxic compounds found
**Philips CA, Valsan A, Theruvath AH, et al. Hepatology Communications 2023;7:e0270 — this is the serious one. 23 patients** with ashwagandha-related liver injury in India; 8 on single-ingredient formulations.
- Cholestatic hepatitis was the commonest presentation
- 5 had pre-existing chronic liver disease
- 3 presented with acute-on-chronic liver failure, and all 3 died on follow-up
This is the pivot for understanding your own risk. The Icelandic and US cases were unpleasant but self-limiting, with zero deaths. Every death occurred in someone who already had chronic liver disease. Pre-existing liver disease is what converts a nasty, recoverable cholestatic illness into a fatal one.
The Netherlands (Lareb, September 2023) adds the detail that matters most for product choice. Four reports, latency 3 weeks to 10 months. The implicated products included Mentalis Stress Trenker (300 mg KSM-66) and Holland & Barrett Ashwagandha KSM-66 (500 mg).
Two of the four were KSM-66 — root-only, aqueous, the extract marketed as the safe one. That fact is why this article is not an advertisement for KSM-66.
LiverTox (NIH) gives ashwagandha a likelihood score of B — "likely cause of clinically apparent liver injury," with around 23 cases in the medical literature as of 2024 and recognition increasing since 2017.
Putting the risk in proportion
Roughly 23 published cases worldwide, plus 11 Australian reports to mid-2026, against 91 listed products in a market where ashwagandha is among the fastest-growing supplements.
But there is no denominator. Spontaneous reporting under-counts by an unknown multiple, and the TGA notes under half its reports came from health professionals. So: the risk is genuinely very rare and almost certainly under-counted, the per-user risk cannot be quantified, and the consequences in one identifiable group are severe.
Warning signs: jaundice, dark urine, pale stools, itching, right upper quadrant pain, nausea, unusual fatigue, loss of appetite. Latency runs from 2 weeks to 22 months — taking it for six months without trouble does not put you out of the window. Prolonged itching with jaundice is the characteristic picture.
Avoid entirely if you have: current or past liver disease, cirrhosis, hepatitis B or C, or fatty liver disease, or if you take other hepatotoxic agents. That isn't a hedge — it's where the deaths were.
Other safety
Pregnancy. The abortifacient claim is traditional, not clinical — the McGill Office for Science and Society is blunt that Denmark's assessment "presents no clinical evidence of ashwagandha as an abortifacient." But RIVM's criticism is better founded: developmental toxicity studies only assessed exposure from gestation day 5 onward, missing the early-pregnancy critical window. That's absence of evidence, not evidence of absence. Australia already mandates a pregnancy warning on certain preparations. Don't take it.
Thyroid medication. See above. This is a thyroid-active substance.
Autoimmune disease and immunosuppressants. The NIH lists a potential interaction with immunosuppressants. The broader immunostimulation concern (Hashimoto's, rheumatoid arthritis, lupus, transplant recipients) is mechanistic and extrapolated rather than supported by case data — but the immunosuppressant interaction is real enough to be in the professional fact sheet.
Sedatives, benzodiazepines, alcohol. Plausible additive CNS depression, consistent with the sleep trials' mechanism. No human interaction study exists.
Diabetes and blood pressure medication. Listed as potential interactions on class grounds. No clinical interaction trials.
Allergens — more practically useful than the nightshade theory. Withania somnifera is in the Solanaceae family, but we found no case reports of cross-reactive allergy. What is on Australian labels matters more: Swisse High Strength contains lactose, milk products and soya beans; Bulk Nutrients contains milk (from KSM-66 processing) and sulphites. KSM-66's manufacturing uses milk, which is a real and poorly advertised issue for dairy-allergic and vegan buyers.
Acute GI reaction. The TGA's 2024 advisory specifically warns of sudden severe vomiting and diarrhoea after one or two doses — distinct from liver injury, and the TGA says so explicitly.
What's on Australian shelves
Cost per trial-equivalent dose (600 mg KSM-66/day, the dose used in Chandrasekhar, Wankhede and Langade):
| Product | Extract | Daily dose | Price | $/day | Trial-equivalent? |
|---|---|---|---|---|---|
| Swisse Ultiboost High Strength | KSM-66, 600 mg, 1 tablet | 600 mg | $44.99 / 60 | $0.75 | Yes, exactly |
| Bulk Nutrients Ashwagandha Capsules | KSM-66, 300 mg ×2 | 600 mg | $39.00 / 120 | $0.65 | Yes, exactly |
| Blackmores Ashwagandha+ | Unnamed extract 150 mg/tab, std. 4.5 mg withanolides, 2 tabs | 300 mg extract, 9 mg withanolides | $14.99 / 60 | $0.50 | No — ~30% of trial withanolide dose |
| Swisse Ultiboost Ashwagandha Calm+ | 30 mg extract ×2, plus passionflower, reishi, magnesium | 60 mg extract | $24.99 / 60 | $0.83 | No — ~10% of trial dose, and dearer per day |
| Swanson Full Spectrum (iHerb AU) | Plain root powder 450 mg | 450 mg | A$13.91 | — | No — unstandardised |
Five things worth noticing:
1. Swisse sells two ashwagandha products whose extract doses differ twentyfold — "High Strength" at 600 mg KSM-66 and "Calm+" at 30 mg per tablet — and the 10%-dose product costs more per day. Same brand, same shelf, same shopper.
2. Blackmores is the honest one, and still falls short. It's the only Australian product we found that states its withanolide content on the label (4.5 mg/tablet) — exactly the transparency this article argues for. That disclosure reveals it delivers 9 mg of withanolides a day against the trials' ~30 mg. Credit for the disclosure; note the shortfall.
3. "Equiv. to dry root 1.5 g" is misleading by being true. A shopper reads "1,500 mg" and thinks that's the dose. The actual extract is 150 mg at 3%.
4. The trial-equivalent options are the cheap ones — 65 to 75 cents a day. Standardisation is not a premium you pay for here.
5. Unstandardised powder still dominates by volume. The top-selling ashwagandha product on iHerb AU is plain root powder with no standardisation at all, outselling every KSM-66 product on the page.
The adulteration problem, which closes the loop
The US Botanical Adulterants Prevention Program issued a bulletin on ashwagandha in October 2025. The principal adulteration: undeclared leaves and stems added to root powders and root extracts, with labels claiming root only. It became acute when COVID-era demand outran production.
And here's the finding that matters most: the bulletin reviewed 22 analytical methods and concluded that "test methods that only assess the presence of withanolides are inadequate and unsuitable" for detecting leaf-in-root substitution — because both roots and leaves contain withanolides.
Separately, trade reporting indicates commercial withanolide content deviates from label claim by two to thirty-five times, with some extracts containing 70–95% uncharacterised material. (We could not trace that figure to a primary analytical source. It is the strongest single number for this argument and therefore the one most worth confirming.)
So a "600 mg ashwagandha root extract" label tells you nothing about dose, little about potency, and — per the adulteration bulletin — nothing reliable about plant part even when it says root, because the routine commercial assay cannot distinguish root from leaf.
If the withaferin-A-in-leaf concern turns out to be real, the testing regime in standard use cannot detect the thing that would matter.
What to do
Don't take it at all if you have any liver condition. That's the one non-negotiable.
If you're going to take it: buy a named standardised extract at a trial-equivalent dose. In Australia that means KSM-66 at 600 mg/day, which costs about 70 cents a day — cheaper than several products delivering a tenth of the dose.
Check the withanolide number, not the milligrams. Around 30 mg of withanolides a day is what the trials used. A 600 mg extract at 1.5% gives you 6 mg.
Ignore the testosterone marketing unless you're an overweight, fatigued man over 40, which is the population the positive trials actually studied — and even there the change stayed inside the normal range.
Expect roughly five minutes off your sleep latency, not transformation.
Stop and see a doctor if you develop yellowing of the skin or eyes, dark urine, itching, unusual fatigue or pain under your right ribs — at any point, even months in.
Frequently asked
Is ashwagandha bad for your liver? It can cause liver injury, but it's rare. The TGA has recorded 11 Australian reports to mid-2026 and about 23 cases appear in the global medical literature. Most cases are cholestatic, appear 2 weeks to 22 months after starting, and resolve within one to five months of stopping. The critical exception is people with pre-existing chronic liver disease — in the Indian case series, all three deaths occurred in that group. If you have any liver condition, don't take it.
Which ashwagandha extract is best? There's no trial comparing them head to head. What matters is withanolide delivery: the trials used roughly 30 mg of withanolides a day, which is 600 mg of KSM-66 at 5% or 250 mg of Sensoril at 10%. An extract standardised to 1.5% at 400 mg gives you about 6 mg. Note that two of the four Dutch liver-injury reports involved KSM-66, so root-only standardisation is not a safety guarantee.
Does ashwagandha raise testosterone? Not meaningfully in healthy men. The positive trials were in overweight fatigued men over 40 or in resistance-training-naïve men, and the clearest result moved testosterone from 310 to 356 pmol/L — within the normal reference range at both ends. No trial shows a clinically meaningful change in a healthy eugonadal man.
Does ashwagandha help with sleep? A little. The pooled effect across five Indian trials was a standardised mean difference the authors themselves described as small, and the best single trial found sleep onset latency improved from about 34 minutes to 29 — roughly five minutes. Quality of life showed no significant improvement in any domain.
Can I take ashwagandha with thyroid medication? Talk to your doctor first. Ashwagandha raised T3 by 41.5% and T4 by 19.6% over eight weeks in people with subclinical hypothyroidism, and there are reports of thyrotoxicosis. If you have Graves' disease or hyperthyroidism, or take levothyroxine or an antithyroid drug, this is a thyroid-active substance taken without monitoring.
Is ashwagandha banned anywhere? Denmark banned it in food supplements in 2023. India mandated root-only in 2026 and banned leaf in food forms. France, Poland and Hungary have issued restrictions or dose caps, and the EU has it under review. Australia has issued two safety advisories and is currently evaluating plant-part evidence, but has not restricted it.
Last reviewed 1 October 2026. Manufacturer funding is noted on every trial where the paper discloses it. TGA advisory details are as reported in the regulator's July 2026 safety update. Prices are Australian retail at the date of review. This page is general information, not medical advice.