We investigated the effect of a 31-d ketogenic diet (KD) on submaximal exercise capacity and efficiency.A randomized, repeated-measures, crossover study was conducted in eight trained male endurance athletes (V˙O2max, 59.4 ± 5.2 mL⋅kg⋅min). Participants ingested their habitual diet (HD) (13.1 MJ, 43% [4.6 g⋅kg⋅d] carbohydrate and 38% [1.8 g⋅kg⋅d] fat) or an isoenergetic KD (13.7 MJ, 4% [0.5 g·kg⋅d] carbohydrate and 78% [4 g⋅kg⋅d] fat) from days 0 to 31 (P < 0.001). Participants performed a fasted metabolic test on days -2 and 29 (~25 min) and a run-to-exhaustion trial at 70% V˙O2max on days 0 and 31 following the ingestion of a high-carbohydrate meal (2 g⋅kg) or an isoenergetic low-carbohydrate, high-fat meal (<10 g CHO), with carbohydrate (~55 g⋅h) or isoenergetic fat (0 g CHO⋅h) supplementation during exercise.Training loads were similar between trials and V˙O2max was unchanged (all, P > 0.05). The KD impaired exercise efficiency, particularly at >70% V˙O2max, as evidenced by increased energy expenditure and oxygen uptake that could not be explained by shifts in respiratory exchange ratio (RER) (all, P < 0.05). However, exercise efficiency was maintained on a KD when exercising at <60% V˙O2max (all, P > 0.05). Time-to-exhaustion (TTE) was similar for each dietary adaptation (pre-HD, 237 ± 44 vs post-HD, 231 ± 35 min; P = 0.44 and pre-KD, 239 ± 27 vs post-KD, 219 ± 53 min; P = 0.36). Following keto-adaptation, RER >1.0 vs <1.0 at V˙O2max coincided with the preservation and reduction in TTE, respectively.A 31-d KD preserved mean submaximal exercise capacity in trained endurance athletes without necessitating acute carbohydrate fuelling strategies. However, there was a greater risk of an endurance decrement at an individual level.