Readability formulae have been used to predict text difficulty, but their construction has been challenged. Previous readability formula designers typically adopted correct responses to reading comprehension questions (RCQs) as dependent variables, but they did not assess the effect of linguistic predictors concerning RCQs in the formula. Thus, this study self-designed a Chinese reading comprehension test, sampled 347 Chinese students in grades 7 and 8 as test-takers, and constructed readability formulae of whole-passage and answer-section models. Correlation and regression analysis showed that the effect of RCQs on the linguistic predictors of readability biased the typical whole-passage model when the model was used to predict text difficulty. We suggest researchers construct updated readability formulae by analysing the effect of RCQs on linguistic predictors of answer-section text to accommodate the present standardised reading tests. We also call for the construction of a readability formula for non-test library reading materials.