Low-carb and keto for cyclists: what the evidence actually shows

Every few years the low-carbohydrate argument comes back round, and it comes back with the same shape. Humans carry tens of thousands of kilojoules of fat and only about two thousand of glycogen. Train the body to burn the abundant fuel, the reasoning goes, and you stop needing to eat every twenty minutes on the bike.

The first half of that is true. Fat adaptation works, and it works faster than most people expect. The second half is where the evidence gets awkward, because the same studies that show enormous increases in fat oxidation also show that riding at race intensity costs you more oxygen for the same speed afterwards.

This piece is about separating those two findings, and about the one version of carbohydrate restriction that has a defensible place in a training year.

What fat adaptation actually does

Take a trained endurance athlete, drop carbohydrate below roughly 50 g a day and push fat to about 78% of energy, and within two to three weeks the muscle rebuilds its metabolic machinery. Fat transport and beta-oxidation enzymes increase. Rates of fat oxidation during exercise roughly double, in some individuals exceeding 1.5 g per minute — a figure that was considered close to a physiological ceiling before this work was done.

That is a genuine, large, repeatable adaptation. It is also not free. The same shift downregulates the enzymes that release energy from carbohydrate quickly, particularly pyruvate dehydrogenase, and it reduces glycogen availability. Fat is the more oxygen-expensive fuel per unit of ATP. At an easy pace that hardly matters. At the intensity where races are decided, it matters a lot.

The study people argue about

The most careful test of chronic LCHF in endurance athletes used elite race walkers in a supervised camp. Three diets, all with matched energy and a hard three-week training block: high carbohydrate, periodised carbohydrate, and a ketogenic LCHF diet. Every group trained identically. Every group improved their aerobic capacity.

Only the LCHF group failed to improve race performance. Their fat oxidation went up dramatically. Their economy went down, and the two findings are causally linked — burning fat costs more oxygen. Because a 10 km race walk is run just under the intensity that the LCHF group could no longer sustain efficiently, the metabolic gain bought them nothing and the economy loss cost them time.

The obvious objection at the time was that three weeks is not long enough to adapt. So the same group ran it again with more athletes and a longer adaptation window, and reproduced the result. That second paper is unusually blunt about it, and it is the reason the high-intensity finding is no longer really contested among researchers. What is contested is how far it generalises past events of that duration and intensity.

Where low carbohydrate may still have a place

Events long enough that nobody goes hard

The economy penalty appears at high intensity. If your event is a twenty-hour unsupported ride at an intensity you could hold while talking, the penalty region in the diagram above is one you never visit, and being less dependent on aid stations has obvious practical appeal. This is the strongest theoretical case for keto-adaptation in endurance sport, and it is also the case with the least controlled evidence behind it. The well-known ultra-runner study measured keto-adapted athletes' metabolism beautifully; it was cross-sectional, so it cannot tell you whether those runners were faster because of the diet or simply the sort of people who stick with an unusual diet. Treat this as plausible and unproven, not as settled.

Metabolic health, where performance is not the goal

Carbohydrate restriction has a real clinical role in glycaemic control and weight management. If you ride bikes and also have a metabolic reason to eat this way, the sensible reading of the evidence is that you will pay something at the top end of your range and probably very little below it. That is a trade you are entitled to make. What you should not do is tell yourself it will make you faster in a road race.

Deliberately training some sessions low

This is the version worth your attention, and it is different in kind from the other two. "Train low, compete high" does not reduce your weekly carbohydrate. It moves it. You put the carbohydrate where intensity is needed, and you allow a specific easy session — a morning ride before breakfast, or an easy ride the evening after a hard afternoon session — to start with low muscle glycogen.

The cell-signalling case for this is solid. Low glycogen amplifies AMPK and p38 MAPK activity and increases PGC-1α transcription after exercise, which is the pathway to mitochondrial biogenesis. The "fuel for the work required" framework formalised it: match carbohydrate to the demand of the specific session rather than to a daily target. The honest caveat is that a 2021 meta-analysis of periodised carbohydrate restriction in trained endurance athletes found the performance effect small and inconsistent. The signal is real; the finish-line consequence is not reliably demonstrated.

What the evidence will and will not support

Two failure modes to watch for

The first is the glycogen mirage. Cut carbohydrate hard and you will lose two to three kilograms in the first fortnight. Almost none of it is fat. Muscle glycogen binds water at roughly three grams per gram, and when the glycogen goes the water goes with it. Riders read the scale, conclude the diet is working, and then cannot understand why their threshold power has fallen by the same proportion. If getting lean is the actual goal, race weight without losing power is the more useful framing.

The second is the slide from periodisation into simple under-eating. "Train low" is a manipulation of timing within a normal energy intake. It is very easy for it to become a smaller energy intake, at which point you are not doing sports science, you are doing low energy availability — and that has a well-documented list of consequences for bone, hormones and immune function. If you are doing fasted or low-carbohydrate sessions regularly, read low energy availability and REDs first and be honest with yourself about which one you are doing.

A practical position

• Default to matching carbohydrate to the work. Around 6–10 g per kg per day in a normal training week, higher on the heaviest days, lower on genuinely easy ones. Protein at 1.6–2.2 g per kg protects lean mass across all of it.
• Keep two or three sessions a week in the low-availability category at most, and make them the easy ones. A pre-breakfast endurance ride is the simplest version; riding fasted covers how to do it without wrecking the rest of the week.
• Never train low before a session that needs intensity. Intervals done on empty are intervals done badly, and a badly executed session trains nothing. If you are using a trainer to hold precise targets — Moveee Indoor will hold them for you and save the power file — you will see the sag in the third block very clearly when you have underfuelled.
• Fuel the event properly regardless of how you train. The capacity to absorb 90 g of carbohydrate an hour is trainable and worth having; see carbs per hour and gut training.
• If you are going to try chronic keto, be clear about what you are buying. You are buying fat oxidation and, quite possibly, a calmer stomach on very long days. You are paying with about 5–8% of your economy in the intensity range where anyone gets dropped.

The uncomfortable thing about this literature is that both camps can quote real data at each other. Fat adaptation is not a myth, and the people who say it does nothing are wrong. But the specific claim that matters to most cyclists — that it makes you faster when the pace goes up — has been tested carefully, twice, and it did not hold. Until someone runs the long-event study properly, that is where the ground is.