Background: Late secondary-level students (those in their final 2 years of secondary school) preparing for the demands of university and careers increasingly face texts written in academic language. Unfortunately, few intervention studies show how instruction can improve their reading achievement. Methods: The 152 student participants (16-17 years old, U.S. Grade 11) were randomly assigned either to small groups receiving scaffolded instruction while reading complex texts or to a business-as-usual control condition. The intervention instruction included tutors supporting students' paraphrasing skills while scaffolding their academic language. Measures included pre-test and post-tests of standardised reading comprehension and tutors' reports of the scaffolds they used in each session. Multi-level regression modelling estimated the causal effect of the intervention (RQ1) and the associations between the types of scaffolds and the comprehension outcome (RQ2). Results: For RQ1, assignment to the intervention significantly predicted a 1.20-point increase in a standardised test widely used for entrance into U.S. universities, which includes a reading comprehension subtest reading scores (d = 0.16). No significant interactions were found by pre-test scores, race, eligibility for school lunch subsidy or learning disabilities. For RQ2, the scaffolds measure explained 11% of the outcome variance, suggesting that scaffolding contributes significantly to comprehension outcomes. Specifically, rereading scaffolds significantly predicted lower post-test scores, while syntax and structure scaffolds predicted higher post-test scores. Conclusions: This intervention presents a model for future interventions with this understudied population. A focus on complex texts with scaffolding designed to account for the complexity of academic language may be effective in supporting late secondary students' reading comprehension. In addition, certain kinds of scaffolding may be differentially effective if educators help students unpack complicated syntax and attend to the structural features of complex texts.