"Generalizability" refers to when trial results are applied to populations that did not participate in the trial. It is important when pre-licensing trial results are used by regulators to authorize the use of a drug in different countries. To "generalize" the results, the trial sample must represent the population from which participants are drawn. They must also ensure that the trial will mimic usual clinical practice. Double-blind, placebo-controlled phase 3 trials are very different from usual practice. Investigators have flagged that samples from phase 3 trials of liraglutide, semaglutide, and tirzepatide excluded one-third of US adults without diabetes who were eligible for weight loss. The authors advised caution on "generalizing" the results to populations excluded from these trials. However, they erroneously used "generalizability" when referring to "representativeness." Similarly, in studies assessing whether real-world populations from China, the United Kingdom, the United States, and Spain were represented in cardiovascular outcomes phase 3 trials in type 2 diabetes with dulaglutide, exenatide, liraglutide, and semaglutide, the authors used "generalizability" in the texts of the articles. The enrolled trial samples were not representative of the real-world population: the term "generalizable" should not have been used. There is widespread confusion among investigators regarding the correct use of the term "generalizability."