Don't know if it's AI slop or not, but for some reason everything to do with nutrition is of low quality and rarely reproducible, and in those cases where it is reproducible, you can find a host of equally reproducible studies pointing in the other direction.

Nutrition is just extremely complex, and monitoring people's diets is very, very hard, and there are huge distortions and incentives to push results one way or another, whether it is businesses (e.g. sugar companies trying to convince people that the real cause of obesity high fat foods and not sugar) or individuals (Walter Willett is on a crusade against animal protein at the Harvard school of public health who is famous for using his position to publish studies in which things like lasagna and pepperoni pizza are classified as meat dishes, and from this reach conclusions that people who eat a lot of meat dishes have worse health outcomes than vegetarians). Basically you have the intersection of something very hard to measure properly even if you are acting in good faith, and then an ocean of bad actors, big money, and ideological people tipping the scales.

At this point, you have to heavily discount all the nutritional studies, unfortunately.