Treating the fact that two things happen together as evidence that one causes the other. The connection may be real — but the direction, mechanism, or existence of causality may be entirely wrong.
Ice cream sales and drowning deaths correlate strongly across the year. One doesn't cause the other — both are driven by hot weather. More seriously: the observation that people who eat breakfast tend to be healthier led to decades of messaging that breakfast "is the most important meal of the day." The relationship was reverse-causal (healthy people have stable enough lives to eat breakfast) and confounded by income.
Our brains are causality-seeking machines. Pattern recognition is our most powerful cognitive tool — but it generates causal stories automatically, and those stories feel like understanding. Correlation data, when presented with sufficient confidence, activates the same neural structures as causal knowledge.
"Studies show X leads to Y" should trigger: "Was this randomized? Is this observational? What's the mechanism? What confounders were controlled for?" Correlation data in headlines almost never answers these questions.