Producing appropriate text alternatives for images in the web is a widely extended accessibility recommendation which, if successfully implemented, facilitates an enhanced web experience for all users, particularly for the visually impaired. Typically, most web accessibility evaluation tools can detect the presence or absence of an alt attribute within the element in a web page. Nevertheless, they rarely perform a deeper linguistic analysis of its content. In the present paper, we introduce Acrolinx, a state-of-the-art controlled-language checker for which we have developed an accessibility-oriented rule set for French, specifically designed to help content authors automatically verify the appropriateness of text alternatives for images.