"Don't exercise late, it'll keep you up" is common advice — but the real picture is more nuanced than the blanket warning suggests, and it's worth being precise rather than just repeating the rule everyone's heard.
The theory behind the warning is reasonable on its face: intense exercise raises core body temperature and heart rate, and both need to come back down before sleep feels natural. Since your body's pre-sleep signal is partly built around a dropping temperature (as covered in an earlier post), exercising right before bed could theoretically work against that mechanism.
What the evidence actually shows, in more detail, is smaller and more individual than the blanket warning implies. Most research finds that moderate exercise, even scheduled in the evening, doesn't meaningfully harm sleep quality for the majority of people — and regular exercise overall, regardless of what time of day it happens, is consistently linked to better sleep than a sedentary lifestyle. The genuine concern tends to be specifically around very intense, high-heart-rate training within roughly an hour of bedtime, which is a much narrower and more specific window than "avoid anything physical after 6pm."
Where individual variation matters most, and where blanket advice tends to fail people: some individuals are simply more sensitive to the stimulating, cortisol-raising effect of a hard training session and genuinely find it harder to wind down for an hour or two afterward. Others train hard at 9pm and are asleep by 10:30 without any issue whatsoever. This is a real case where generic advice doesn't apply evenly across people, and paying attention to your own actual experience is more reliable than following a general rule that may not describe you at all.
The practical takeaway: if you already train in the evening and sleep just fine, there's no strong reason to change that based on this alone. If you train hard late in the day and consistently struggle to fall asleep afterward, that specific combination is worth testing directly — shift the session an hour or two earlier for a week or two and see if it actually changes anything, rather than assuming the connection exists without checking.