What to eat: what matters and what doesn't
Step 4 of 6. "What should I eat to lose weight?" is the question the whole diet industry lives on. The trials have an answer, and it reorganises everything else on this site.
The finding that reorganises everything
When people are randomly assigned to lower-carbohydrate, lower-fat or mixed patterns with a similar energy deficit, they lose similar amounts of weight at twelve months. Three pieces of evidence carry that sentence:
- A network meta-analysis of 121 randomised trials with nearly 22,000 participants compared fourteen named diet programmes. At six months most produced modest weight loss and better blood pressure; at twelve months the differences between them had largely disappeared.
- A one-year trial of 609 adults compared a healthy low-fat diet with a healthy low-carbohydrate one. The two groups lost the same amount on average, and neither genotype nor insulin secretion predicted which diet suited whom — the hypothesis the trial was built to test came out negative.
- A two-year trial of 811 adults assigned four different macronutrient compositions. Weight loss differed between the groups by less than half a kilogram, and all groups slowly regained after the first year.
The current clinical guidelines say the same thing in their own words: several patterns can achieve similar weight reduction, and the choice should be individualised. The practical translation is blunt — the pattern you keep is the one that works.
The two things that matter inside any pattern
Protein. This is the one macronutrient with a consistent effect during weight loss. A systematic review of trials in adults with overweight found that higher protein intake helps preserve muscle, with the signal strengthening above roughly 1.3 grams per kilogram of body weight. Clinical guidance for active weight loss puts the working range at 1.2 to 1.6 grams per kilogram, with something like thirty grams per meal for satiety. For comparison, the European reference intake for a healthy adult who is not losing weight is 0.83 grams per kilogram — the maintenance norm, not the target during a deficit.
Fibre. Fibre is what makes a smaller amount of food feel like enough. The general planning target is fourteen grams per thousand kilocalories, which for most deficits works out around twenty-five grams a day and up. Vegetables, legumes, whole grains and fruit get you there; a pattern that cuts all of them tends to run short.
What matters much less than people think
- Exact macronutrient ratios. Inside the wide healthy-population ranges, the ratio is a preference.
- Meal timing. A tool for keeping a deficit, for some people; not a lever on the deficit itself. A later page in "Check a claim" looks at the evidence.
- "Clean" versus "processed" as a rule. Less processed food is usually easier to eat less of and better for you in other ways; it is not a weight-loss mechanism on its own, and turning it into a rule is one more way to fail.
How to choose
Since adherence decides, choose on the questions that predict adherence for you:
- Do I cook, or do I need something that survives a canteen and a takeaway?
- Do I eat with a family that will not change with me?
- How often do I eat out or with friends?
- What can I afford every week, not just this week?
Every way-of-eating page on this site scores those four things, alongside the evidence. If you want a named pattern, the comparison is in Ways of eating; if you do not, "eat mostly whole foods, hit your protein, get your fibre, keep the deficit" is a complete answer.
What this page does not claim
That any food is forbidden. That a named diet is best. That protein above the range adds anything. Anything about supplements — none of the sources above needed one.
In the app, the generated menu has to hit the day's energy target and the balanced policy, and it is checked deterministically before it reaches you — not by the model's own say-so. Named patterns are a future goal, not a promise.