The research on exercise and depression is better than the sceptics say and far weaker than the headlines claim. Read it closely and it points somewhere specific: the thing that works may be the part we cannot give away for free.
If you are struggling with your mental health right now, please talk to your GP or a crisis line in your country. Nothing below is a reason to delay treatment or to change medication you are taking.
“Exercise is as effective as antidepressants.” “A brisk walk beats medication.” You have read these headlines. They come from real research — a 2024 network meta-analysis in the BMJ pooling 218 studies and 14,170 participants, and a 2023 overview covering 97 systematic reviews and over a thousand trials.
The research is real. The headlines are not what it says. And the gap between them matters, because the advice that falls out of the headline version — go for a walk, it works as well as pills — is both unsupported and, for someone in the middle of a depressive episode, quietly cruel.
The effect shrinks every time the methods get stricter
Start with the pattern that runs through this entire literature. Exercise trials cannot be blinded — you always know whether you have been jogging — and the effect behaves accordingly.
The 2013 Cochrane review found an effect size of −0.62 across 35 trials. Restricted to the six trials with adequate allocation concealment, intention-to-treat analysis and blinded outcome assessment, it fell to −0.18, and was no longer statistically significant.
A 2017 analysis in BMJ Open of 35 trials found −0.66 overall. Among the four low-risk-of-bias trials: −0.11, null. Its authors concluded that trials at less risk of bias “suggested no antidepressant effects of exercise.”
A 2025 review pooling 11 meta-analyses and 16,255 participants found −0.61. After correcting for publication bias by trim-and-fill: −0.24, non-significant. Its test for funnel-plot asymmetry returned t = −9.35, p < 0.0001.
The newest Cochrane update, published in January 2026 and covering 73 trials, found −0.67 overall and −0.46 among the seven high-quality trials — still favourable, but with a confidence interval whose upper bound is −0.04, and rated low-certainty evidence. Long-term follow-up was not statistically significant.
And the BMJ paper behind the biggest headlines states plainly that of its 218 studies, exactly one met Cochrane’s criteria for low risk of bias. It rated its own confidence as low for walking and jogging, and very low for everything else.
None of this means exercise does nothing. It means the honest estimate is smaller and shakier than “as good as antidepressants,” and that reasonable methodologists genuinely disagree about whether a specific antidepressant effect has been demonstrated at all.
Roughly half the benefit shows up in the placebo arm
There is one way to get at the blinding problem: the rare three-arm trials that include exercise, a credible placebo exercise condition, and a control.
A meta-analysis of nine such studies covering 661 participants found the exercise effect at d = 0.37 — and the placebo exercise effect at d = 0.20. The authors put it directly: the placebo effect accounts for approximately half the observed psychological benefits of exercise training.
Expectation, attention and the structure of turning up somewhere three times a week are doing a substantial share of the work.
The trial that tested the actual advice — and failed
Almost every favourable estimate in this field comes from supervised, structured programmes: typically three sessions a week for ten to twelve weeks, often led by a physiotherapist or exercise physiologist.
That is not what “just go for a walk” means. And there is a trial that tested the thing we actually tell people.
The TREAD trial randomised 361 depressed adults in UK primary care to usual care, or usual care plus a facilitated physical activity intervention. The intervention worked as designed — it significantly increased physical activity, odds ratio 2.27. On depression, the between-group difference at four months was −0.54 on the Beck Depression Inventory, with a confidence interval from −3.06 to 1.99. Nothing. No reduction in antidepressant use either.
Successfully getting depressed people to move more did not make them less depressed.
The symptom being treated is the symptom blocking the treatment
This is the part the cheerful version of this advice ignores entirely, and it is not a rhetorical objection — it is the proposed mechanism, running backwards.
A 2024 review in Translational Psychiatry argues that exercise’s antidepressant action works partly through dopamine transmission and effort-based decision-making for reward. Which is precisely what depression degrades. Anhedonia and reduced willingness to exert effort are not failures of character; they are diagnostic features of the illness. The intervention is effortful, and the condition is defined in part by the collapse of the capacity for effort.
The trial data reflects this. A meta-analysis of dropout across 40 trials and 1,720 participants found that worse baseline depressive symptoms predicted higher dropout (β = 0.0409, p = 0.04). The people most in need of the benefit are the least likely to complete the programme.
The same analysis found what protected against dropout: supervision by a physiotherapist (β = −1.2029, p = 0.008) or an exercise physiologist (β = −1.3396, p = 0.01). Not willpower. A named professional expecting you.
What the guidelines actually say
Worth knowing, given how often exercise is presented as a frontline alternative.
NICE guideline NG222 lists group exercise sixth of eleven options for a new episode of less severe depression — behind guided self-help, group CBT, group behavioural activation, individual CBT and individual behavioural activation. For more severe depression, group exercise ranks last, tenth of ten.
The VA/DoD clinical practice guideline suggests exercise “as an adjunct” — strength of recommendation: weak for.
No guideline body recommends exercise as first-line monotherapy for moderate or severe depression.
Prevention and treatment are not the same question
One more conflation worth separating. A 2022 JAMA Psychiatry meta-analysis of 15 cohort studies and 191,130 people found a dose-response relationship: adults getting about half the recommended activity had 18 per cent lower risk of developing depression, and those meeting it, 25 per cent lower. It estimated 11.5 per cent of cases might have been prevented.
That is observational, it is about not becoming depressed, and it says nothing about treating an episode already underway. The genetic evidence meant to shore up causality is itself unstable — Mendelian randomisation gave an odds ratio of 0.74 in 2019 and 0.92 in 2023 once sample overlap and winner’s curse were corrected.
Where this leaves things
Exercise is worth doing. It is good for a great many things, the trial dropout rates are no worse than control conditions, and for mild depression a structured programme is a legitimate option that guidelines do list.
But the specific claim that has been circulating — that a walk is a substitute for treatment — is not what the evidence shows. What the evidence shows is that supervised, structured, professionally-delivered programmes produce a modest effect that shrinks as bias controls tighten, that about half of it is reproduced by a placebo exercise condition, that the one large pragmatic trial of the real-world version found nothing, and that severity itself predicts who drops out.
Which suggests the active ingredient may be the expensive part — the structure, the supervision, the human being expecting you on Tuesday. “Just go for a walk” is what remains after you have subtracted everything that made it work.
If you are depressed and cannot make yourself exercise, you have not failed at an easy thing. You are experiencing the illness, and the research says so.
This article is general information, not medical advice, and it is not a guide to treating depression. Do not stop or reduce antidepressant medication on the basis of anything here — nothing in this literature supports discontinuation. If you are depressed, speak to your GP or a mental health professional. If you are in crisis, contact your local emergency services or a crisis line now.
