Why the science is hard

And why headlines conflict.

Method problems

Long randomised diet trials are expensive and rarely feasible, so most evidence is observational. People misreport what they eat, and diets correlate with income, education and exercise.

Confounding is severe: healthy eaters differ in many other ways.
Effect sizes for single foods are usually tiny.

Reading claims

Ask whether it was randomised, how long it ran, how diet was measured, and how large the effect was in absolute terms. Most reported findings shrink under those questions.