How to Lower Cortisol: Meditation or Yoga. Cortisol May Not Be Why They Help.

Meditation lowers measured cortisol: a little, on average, and not in every group that has been studied. Randomised trials say so, and the two reviews this page started from used the hard version of the test — every trial compared against another activity rather than against doing nothing. [1][2] Yoga does too, most clearly in people who were under strain to begin with. [2][10]

Whether that is why anyone feels better is a different question. It has been tested only a handful of times, and the tests do not agree.

Cortisol was not the only thing those trials measured, and not the only thing that moved. And a “cortisol detox” — a phrase that draws about seventy-four thousand US searches a month on Google, by the figure we pulled on 8 August 2026 — is not a procedure that exists.

The test was the hard kind. The count was not 87.

Two meta-analyses, published in 2017 by the same research team, took in 45 randomised controlled trials of meditation and 42 of yoga. [1][2] This page used to add those together and put 87 trials in its headline. That was the wrong number to put there. The reviews took in any trial that measured any of a list of stress markers — blood pressure, heart rate, inflammation, blood fats, cortisol — and at least five trials sit in both. Seven of the meditation trials are in that review’s cortisol analysis, 212 people in all. Eight of the yoga trials are in the other’s. [1][2]

We read those counts off the charts inside the two papers that list each trial and its size (forest plots), which the publisher shows openly. The papers’ text is behind a paywall and we have not read it, so what this page says about these two reviews comes from their summaries and their charts.

And they did the hard version of the test. Every trial included was compared against an active control: another activity, not a waiting list. That matters enormously. Compare meditation to doing nothing and you measure the effect of having something to do. Compare it to a different structured activity and you measure the meditation.

The researchers were explicit about why, noting that studies in this field “commonly fail to include an active control group.” [1]

Both found cortisol went down. In the meditation review the seven trials (they measured cortisol in blood, by a later review’s account [5]) together showed a fall of about half a standard deviation, the statisticians’ measure of how much people differ from one another to begin with: 0.52, on a scale where 0.2 is conventionally small, 0.5 medium and 0.8 large. The range the true average probably sits in runs from 0.24 to 0.80, which is wide, as you would expect from 212 people. [1] Going by the trials’ titles, five of the seven were in people with a medical condition: HIV, diabetes, heart failure, depression in old age. The yoga review pooled cortisol by time of day and found it lower on waking, in the afternoon and in the evening, with no clear difference half an hour and an hour after waking, in mid-morning, or in how steeply it fell across the day. Its largest single comparison had 386 people, most of them breast cancer survivors or people with metabolic syndrome, the cluster of high blood pressure, high blood sugar and a large waist. [2]

Look at the other rows

This is the part that changes what the answer means.

Cortisol was not the only thing that moved. With every form of meditation pooled, blood pressure fell, resting heart rate fell, triglycerides (a blood fat) fell, and so did two blood markers of inflammation: C-reactive protein and a signalling protein called TNF-alpha. [1] In the yoga trials: cortisol at three times of day, blood pressure at rest, resting heart rate and, by the review’s own summary, fasting blood sugar, cholesterol and LDL, the kind of cholesterol doctors want lower. [2]

That reads like everything moving together. It did not. In the meditation review the inflammation marker IL-6 did not change, nor did cholesterol, and heart rate tracked through the day did not fall. [1] The yoga review found no clear change overall in the inflammation markers it pooled, and its summary lists a fall in one kind of heart-rate variability (the beat-to-beat variation in the pulse) that we cannot interpret without the full paper. [2] And each marker was pooled from its own handful of trials, mostly different ones: of the seven meditation trials in the cortisol analysis, only one is also among the eleven in the resting blood-pressure analysis, and the same one among the nine for resting heart rate. [1]

Which means nothing here shows cortisol is the lever. It is one dial among several that moved, mostly read in different people, and no trial was built to ask whether the cortisol dial drives any of the others. You cannot tell from this whether lowering cortisol did anything, or whether cortisol simply fell along with other things because the person was less stressed.

So how much is the prize worth?

If you are chasing cortisol specifically, it is worth knowing how loosely it connects to the things people actually want changed.

Across 122 cohorts and 26,527 people, the correlation between long-term cortisol measured in hair and body mass index was 0.10; against waist size, in 24 cohorts and 11,006 people, it was 0.11. [3] A correlation runs from 0, no relationship at all, to 1, where one number tells you the other exactly. Squared, 0.10 is about one percent of the variation. And it is a snapshot: it compares people at one moment, so it cannot say whether cortisol adds weight, weight raises cortisol, or something else moves both. The authors say so themselves: “causality is not yet proven”. [3] (A cohort is one group of people studied together. The whole review took in 146 of them and 34,342 people; the figures here are for the cohorts that reported each measurement.) We took that apart in our verdict on cortisol face and cortisol belly.

Hair cortisol, the long-term measure behind that 0.10, is one that meditation trials have not moved. Two trials and 223 people in the 2021 review below, three comparisons and 306 employees in a 2026 one: no change that either could detect. [5][8]

And the daily cortisol curve — the thing “rhythm resets” are sold against — relates to health outcomes across 80 studies at an average correlation of 0.147, which its own authors call small. They leave open whether a flat curve is a mechanism that causes poor health, an effect of it or a marker of something else, and argue it may both reflect and contribute. [4] That one is covered in our cortisol rhythm verdict.

So: an intervention that moves a number, on average and not in every form or every group, and a number that is weakly attached to the outcome. The intervention still looks good. The number is just a bad way to judge it.

A detail that got flattened everywhere else

Worth pulling out, because summaries of this research lose it, and this page’s first version got part of it backwards.

The meditation analysis split practices into three kinds, and they did not all do the same thing. Focused-attention practice, where you hold your attention on one thing, such as the breath or each step, and bring it back when it wanders, lowered cortisol: three trials and 102 people, two of walking meditation and one of a technique called Integrated Amrita Meditation. Mantra practice of the Transcendental Meditation kind did not, in its three trials and 72 people. Open-monitoring practice, the mindfulness style of watching whatever comes up without reacting to it, had a single cortisol trial; what it lowered was resting heart rate. Each kind lowered systolic blood pressure, the top number: the first two at rest, the mindfulness style only when pressure was tracked through the day. [1]

Pool all seven trials and cortisol falls, but that average leans on the three focused-attention trials. [1] “Meditation lowers cortisol” is therefore true of the average, and not equally true of whichever practice you happen to have downloaded. In 2017 the mindfulness style had one cortisol trial to its name. It has many more now, and they are next.

How big, and in whom? The newer reviews disagree

The 2017 reviews are not the last word. Reviews published since have pooled the cortisol results of many more trials, and they report sizes on the same yardstick as above: the gap between the group doing the practice and the comparison group, counted in standard deviations. Each is an average across groups of people. None is a forecast for one person.

Meditation of any school, 2021. In ten trials that took blood samples (395 people) the fall was 0.62, and the range the true average probably sits in runs from 0.22 to 1.02. On a closer look it was clear only in patients with a physical illness. In 21 trials that took saliva (1,163 people) it was 0.18, with a range from slightly below zero to 0.40, so “no effect” is still on the table; the authors call that result not robust and found signs that studies with nothing to report were missing. [5]

Stress-management programmes in healthy adults, 2024. Across 58 trials and 3,508 healthy adults the fall was 0.28 (range 0.17 to 0.40). For mindfulness and meditation programmes it was 0.35. For the group of practices that holds yoga it was 0.13 and could not be told apart from zero, although the review could not show that the kinds of programme really differed from one another. And it found the fall in adults who were not under particular stress and no clear fall in those who were: the reverse of the 2021 review, as its authors note, though the gap between those two groups was itself small enough to be chance. [6]

Yoga and other exercise in people under strain, 2025. A review of 44 trials and 3,284 people who scored high for depression, anxiety or stress set kinds of exercise against one another. Yoga, the subject of twenty of the trials, came out first: against the comparison groups the fall was 0.59 (range 0.28 to 0.90), and the reviewers graded that result high-certainty, the only one they did. Two things to hold against it. The comparison groups were waiting lists, usual care, no programme at all or a different kind of exercise, which is not the hard version of the test the 2017 reviews used. And most of the spread between the trials’ results was more than chance would produce: an I-squared of 91% across the 17 that set yoga against a comparison group, and 86% across the whole analysis. That is a measure of how much of the spread luck does not explain, not of which way the trials pointed, and the review does not print its trials one by one, so we cannot tell you how many of them found cortisol falling. The other kinds of exercise it compared showed smaller or unclear changes (high-intensity intervals tended, if anything, to raise cortisol), and cortisol measured in hair did not change. [10]

Mindfulness-based stress reduction, the standard mindfulness course, September 2026. The newest review pools 39 studies, not all of them randomised trials, and 2,059 people. Against comparison groups of any kind the fall was 0.38 (range 0.21 to 0.56). Against active comparison groups, the hard version of the test, it was 0.29 (range 0.04 to 0.54), from eleven comparisons. The authors say the results vary from study to study more than they can explain. We have read this one only as its published summary, with the data its authors posted; the paper itself is behind a paywall. [7]

Meditation at work, April 2026. A review of 132 trials in employees found that people taught meditation reported less stress, anxiety and depression afterwards. In the ten comparisons that also measured cortisol (778 people), it had not changed by more than chance could explain when the programmes ended. Four comparisons measured again within three months and found a small fall, 0.24, which cleared the line researchers use for “probably not chance” in the review’s main analysis and slipped just the wrong side of it when analysed as change from each group’s starting level. The reviewers graded the cortisol evidence low. [8]

So the direction mostly holds, and that is what the rating rests on. What does not hold still is how much, measured how, and in whom: patients in one review and unstressed healthy adults in another, blood but not saliva in one and no difference between the two in another, and in working adults nothing detectable by the end of the course. That is why this page says “a little, on average” where it used to say “you can”.

What we are not going to tell you

A dose. The yoga authors state plainly that “the reported interventions were heterogeneous” [2] — different practices, lengths and frequencies. Neither summary gives a schedule, and the later reviews do not settle one: one found bigger falls in blood cortisol once a programme passed about twenty hours in total, another found the length of a programme made no difference. [5][6] So we give none either.

An effect size of our own. The reviews above publish theirs, and they do not agree. We could average them into one tidy number. We are not going to, because an estimate printed next to real numbers becomes a real number by morning.

A supplement. Ashwagandha is one that is sold for this, and it has its own verdict, including what it did and did not move.

Anything about sleep, other exercise or diet. This page checked two practices, meditation and yoga. Whatever else is recommended for cortisol is not rated here. The 2025 review above is the only source on this page that looked at other kinds of exercise, and we have not checked that literature ourselves.

The verdict

Meditation and yoga measurably lower cortisol, by a small to moderate amount on average. That is what our Supported rating covers. For meditation it rests on pooled randomised trials in three reviews by three different teams, including against active controls, and a fourth review, of the standard mindfulness course, that points the same way. [1][5][6][7] That is a better standard of evidence than most wellness claims ever face. The limits are real: the reviews share some of their trials, so they are not separate confirmations; the first rests on seven trials; one found no clear fall in saliva samples; and a fifth review found none in working adults by the end of a course. [5][8]

For yoga it rests on two reviews, and on who was in them. The 2017 one, eight trials against active controls, at least half of them in people with a medical condition or recovering from cancer, found cortisol lower at three of the seven times of day it checked. [2] A 2025 one found the clearest fall of any kind of exercise for yoga, across twenty trials in people with raised scores for depression, anxiety or stress, against easier comparison groups and with results that varied widely. [10] And the 2024 analysis of healthy adults found no clear fall for the group of practices yoga sits in. [6] So: clearest in people under strain, and not shown in people who are well.

Whether lowering cortisol is why they help is not settled. The reviews above do not test it. A few single trials have, by checking whether the change in cortisol accounts, statistically, for the benefit (researchers call that a mediation analysis), and they disagree. In 228 adults with chronic low back pain, randomised to yoga classes, stretching classes or a self-care book, cortisol did not qualify as an explanation for what yoga did, among the 82 people in that comparison who gave saliva. [11] In 118 older adults, eight weeks of yoga against stretching improved working memory, and the trial’s summary says the gain was carried by a smaller stress response, salivary cortisol included. That trial is one of the eight pooled for yoga above, and we have read only its summary. [12] A 2015 systematic review found five yoga studies that had tested any explanation this way; by its summary, salivary cortisol was one of four that held up, and its authors asked for larger, randomised, better-controlled studies. [13] So: tested a little, in yoga, with answers that point both ways. What the reviews do show is that cortisol was one of several measures that moved, mostly in different trials, and that in working adults reported stress fell while cortisol, at the end of the course, had not detectably changed. [1][8]

Here is the practical version. If you took up meditation and felt better, there is trial evidence on your side, and it does not run through a hormone. A review of 47 trials, each against an active control, in patients and in people under stress, found that mindfulness programmes brought small improvements in anxiety and depression, about a third of a standard deviation after eight weeks. It found no evidence that they beat other active treatments such as drugs, exercise or behavioural therapy, and its evidence on body weight, which it graded low, pointed to no effect. [9] It left yoga out, and no source on this page measures how people feel after yoga. If meditation was sold to you as a way to move a hormone, you were sold the least informative reason available — and if you are buying something to move that hormone, you are paying for the dial rather than the driving.

None of this is about cortisol that is medically too high, as in Cushing’s syndrome, or about people taking steroid medicines. That is a question for a doctor, and what doctors look for is on our page about high cortisol symptoms. We read the research; we are not your clinician.

Related: cortisol face and cortisol belly · cortisol rhythm and fat loss · ashwagandha benefits · box breathing · all the habits we have checked