Supplement Insider Evidence-based wellness
Ingredients / Ashwagandha
Moderate evidence overall 24 articles · reviewed 2 Sep 2026

Ashwagandha

Withania somnifera

An adaptogen with a genuinely decent trial record for stress and sleep, and a much thinner one for everything else it's sold for.

What the evidence supports

Moderate Stress & cortisol Several randomised trials show lower perceived-stress scores over 8 weeks. Most are small and industry-funded.
Moderate Sleep quality Improves sleep-onset latency and self-rated quality, most clearly in people with insomnia.
Limited Strength & recovery A handful of resistance-training trials with small samples and inconsistent outcomes.
Limited Testosterone Changes reported are small, in specific populations, and not replicated in healthy men.
Dose used in trials
300–600 mg

per day of a root extract standardised to 5% withanolides, taken for 8 weeks or more.

Safety
Generally tolerated

Mild GI upset and drowsiness are the common reports. Rare liver-injury case reports exist.

Ask a doctor first if
Four cases

Pregnancy, thyroid medication, immunosuppressants, or sedatives.

The trials, individually

Open any one for sample size, funding and what it actually measured.

Chandrasekhar et al. · 64 adults · stress
Double-blind, placebo-controlled, 60 days, 300 mg twice daily. Perceived Stress Scale fell markedly versus placebo; serum cortisol also dropped. Small sample, single site, extract manufacturer supplied the material.
Langade et al. · 60 adults · insomnia
Ten weeks, 300 mg twice daily, actigraphy plus sleep diaries. Sleep-onset latency and efficiency improved, with the larger effect in the diagnosed-insomnia subgroup.
Wankhede et al. · 57 men · strength
Eight weeks alongside resistance training. Bench-press and leg-extension gains exceeded placebo, but the training programme was not individually standardised and dropout was uneven.
Systematic review · 12 trials · anxiety
Pooled effect favours ashwagandha but heterogeneity is high and most included trials carry a moderate-to-high risk of bias. The reviewers' own conclusion is "promising, not settled".

Related reading