Early or indirect evidence. Worth trying, not worth promising.
Probably a little, at most, and possibly not at all. Magnesium is one of the most-bought sleep supplements in Australia and one of the least-supported things in our library. We grade it the way the reviews read, not the way it sells.
If you are about to buy magnesium for sleep, the useful thing to know is not which form to get. It is how strong the evidence is that any of them will do much — and that is where almost every page on this subject goes quiet.
What the trials found
Systematic reviews of magnesium for sleep rate the evidence low to very low certainty. A handful of small studies found people fell asleep somewhat sooner. Total sleep time was no different. That is the whole finding, and the honest summary of it is “small effect at most” rather than either “proven sleep aid” or “useless”.
Low certainty is a specific claim, not a hedge. It means the studies were small enough, or built loosely enough, that a future trial could easily land somewhere else. Anyone telling you magnesium is established for sleep is describing a literature that does not exist yet.
Food first, and why that is not a dodge
Nuts, seeds, beans, leafy greens and whole grains carry magnesium, and eating them is the version with no label to read, no interaction to check and no purchase to make. It is also the version most of the supporting argument for supplementation quietly rests on — low intake being the reason to top up.
Two things with far better evidence
If the goal is falling asleep more easily, our library rates both of these higher than magnesium, and both are free:
- A wind-down, rated B. Screens away and slow breathing before bed. Long, slow exhales shift the nervous system toward rest, and a consistent pre-sleep routine is one of the most reliable sleep-quality improvements there is.
- A caffeine cutoff, rated A. A meta-analysis of twenty-four studies found a coffee within about nine hours of bedtime cut the night by roughly three quarters of an hour. A pre-workout product needs thirteen. This is the single biggest sleep lever most people have not pulled.
Both are in IntentNorth, scheduled at real times in your week rather than left as advice. So is magnesium — with its D on it.
If you try it anyway
That is a reasonable thing to do with a D. Thin evidence means untested, not disproven, and a cheap, low-risk thing with a small possible upside is a fair experiment. Give it a fortnight, keep everything else the same, and judge it on your own nights rather than on the packet.
What we will not tell you
An amount, or a form. Neither is ours to give: the trials are too thin to justify a number and any figure printed here would outrank a label and a pharmacist for somebody who read it, which is exactly backwards. Every supplement entry in the library works the same way — what the public evidence shows, graded, and no dose.
Before you try it
A kidney condition changes how the body clears magnesium and makes this a doctor’s call. Loose stools are the usual sign of too much. Education, never advice: nothing here tells you to take anything, and no amount on this card outranks a label or a professional. Talk to a doctor or pharmacist before starting anything — especially if you take any medication, because interactions are real; if you are pregnant or breastfeeding; if you have a kidney or liver condition; or if you have any history of disordered eating.
IntentNorth is an iPhone app that puts practices like this one into your actual week — at real times, around what you already have on, with the grade on each one. It changes the week when your week changes.
How this is graded. Every practice in IntentNorth carries a letter from A to E for the strength of the research behind it, shown on the practice itself rather than in a footnote. All 204 are published, and 122 of them are rated C or weaker. Read the whole library.
Educational information, not medical advice. Nothing here is a diagnosis or a personal recommendation, and none of it replaces a conversation with your doctor or pharmacist.
Last reviewed September 2026