Who and how many?
Start with the size and type of the study. A result from ten people is a hint, while a result from ten thousand is more reliable. Studies on mice, cells or computer models do not automatically apply to humans. Good articles tell you who was studied, and if it is missing, that is worth noting.
Association is not cause
Many health stories report that people who do one thing are more likely to have some outcome. That is an association. It does not show that the first thing caused the second, because other factors may explain both. Wording such as 'linked to' or 'associated with' is the careful way to say this, whereas 'causes' needs much stronger evidence.
Absolute and relative risk
Saying a habit 'raises your risk by 30 percent' sounds alarming. If the risk was one in a thousand it is now 1.3 in a thousand. Look for the actual numbers behind any percentage.
Has it been checked?
Peer review means other experts examined the work before publication. It is not a guarantee, but results announced only in a press release or a preprint are less established. One study alone rarely settles anything, and results are much more convincing when other teams reproduce them.
Who paid and who is quoted
Funding does not automatically spoil a study, but it is worth knowing. Good coverage also quotes independent experts who were not involved, who can say whether the finding is a small step or a breakthrough.
Be wary of miracle language
Words such as 'breakthrough', 'cure' and 'game-changer' are red flags. Real progress is usually gradual and cautious. When a headline sounds too good or too frightening, read the study's own summary, called the abstract, which is usually free.