This study aimed to determine the short-term effect of two isocaloric diets differing in the ratio of protein–carbohydrate on melatonin levels, sleep, and subsequent dietary intake and physical activity in healthy young men. Twenty-four healthy men took part in a crossover design including two sessions of three days on isocaloric diets whether high-protein, low-carbohydrate (HPLC) or low-protein, high-carbohydrate (LPHC) followed by 24-h free living assessments. Sleep was measured by ambulatory polysomnography pre-post-intervention. Melatonin levels were assessed on the third night of each session on eight-point salivary sampling. Physical activity was monitored by accelerometry. On day 4, participants reported their 24-h ad-libitum dietary intake. LPHC resulted in better sleep quality and increased secretion of melatonin compared to HPLC. A significant difference was noted in sleep efficiency (p < 0.05) between the two sessions. This was mainly explained by a difference in sleep onset latency (p < 0.01) which was decreased during LPHC (PRE: 15.8 ± 7.8 min, POST: 11.4 ± 4.5 min, p < 0.001). Differences were also noted in sleep staging including time spent on REM (p < 0.05) and N1 (p < 0.05). More importantly, REM latency (PRE: 97.2 ± 19.9 min, POST 112.0 ± 20.7 min, p < 0.001) and cortical arousals (PRE: 7.2 ± 3.9 event/h, POST 8.5 ± 3.3 event/h) increased in response to HPLC diet but not LPHC. On day 4, 24-h ad-libitum energy intake was higher following HPLC compared to LPHC (+64 kcal, p < 0.05) and explained by increased snacking behavior (p < 0.01) especially from carbohydrates (p < 0.05). Increased carbohydrates intake was associated with increased cortical arousals.