This project explores the assessment of academic writing for U.S. students learning English as a second language. Through the analysis of 1200 student responses to the writing component of a large-scale standardized test of academic English language development, the study explores how students in grades 3, 6, and 9 at four different score levels use language from task input in their responses. Drawing from research literature about integrated tasks and source-based writing (Shi, 2004; Weigle & Parker, 2012), the study adapts methodologies for analyzing student responses and applies these to a K-12 assessment context. Assessment tasks in the study are described as input-rich tasks and present students with text and graphic prompts in order to elicit responses that reflect academic language proficiency. Results suggest that while a large portion of language in student responses comes directly from the task input, extensive borrowing of longer strings of text is relatively rare across grade and score levels. Clear patterns of language use differentiate students by score level.