Metacomprehension as reflected in judgements of one's learning is crucial for self‐regulated study, yet their accuracy is often low. We investigated text difficulty as a constraint on metacomprehension accuracy in text learning. A total of 235 participants studied a 10‐section expository text and afterwards took a knowledge test. They made judgements of learning after each section. Sections were of high, medium or low difficulty; we manipulated between participants the order of difficulty levels across sections. In blocked orders, texts in each block (sections 1–4; sections 5–6; sections 7–10) were of the same difficulty level. In mixed orders, difficulty varied throughout the learning unit either from easy to difficult or from difficult to easy. Our general tenet was that orders would trigger different extents of experience‐based processing and thus influence metacomprehension accuracy to different degrees. As hypothesized, accuracy was higher for blocked difficulty orders. Late‐section judgement magnitude decreased more strongly in the blocked groups. At the same time, late‐section judgement accuracy was higher in the blocked group. We discuss implications and limitations of the influence of fluctuations in text difficulty on judgements of learning accuracy together with some avenues for further research. Lay Description What is already known about this topic? Metacomprehension as reflected in judgements of one's learning (JOL) is crucial for effective self‐regulated study. JOL accuracy is often low. Text difficulty appears to bias metacomprehension in topically non‐coherent series of texts. What this paper adds? This paper establishes text difficulty as a constraint on metacomprehension accuracy in topically coherent texts. This paper explains text difficulty effects as different relative weights of experience‐based and theory‐based processing, respectively. Implications for practice and/or policy: Taking systematic text difficulty effects into account may help avoid unintended adverse effects of difficulty order on metacomprehension accuracy. Arranging difficulty levels block‐wise may enhance experience‐based processing and thus judgement accuracy. Additionally, with topically coherent texts, increasing difficulty across texts benefits performance in a later knowledge text.