Since the advent of the Internet, access to information knows no boundaries. The Web is home to a complex array of (hyper)texts and multimedia content that are easy to interact with when a computer screen and a mouse are available. Nevertheless, when digital information can be only accessed if transformed into audio output, interaction becomes troublesome. Blind users, who rely on assistive technologies to perceive the Web, experience numerous challenges when accessing visual content, such as images. Web accessibility guidelines prescribe that images must be complemented with an appropriate description, that will act as a textual replacement when they cannot be displayed (Caldwell et al. 2008). Yet, studies have demonstrated that these text alternatives are often missing and that, when present, they are not correctly implemented (Hanson and Richards 2013). In the context of the multilingual Web, image accessibility heavily depends on the web localiser's ability to identify image text alternatives as translatable elements within an HTML document, and to effectively assess their appropriateness. Accomplishing these two steps is key to deliver a fully functional target web product. While scholars have already pointed to web accessibility as a component of web localisation quality (Jiménez-Crespo 2013), there remains a paucity of empirical evidence on how it is addressed during a localisation task. This paper presents the results of an image accessibility experiment involving 28 localisers with and without previous knowledge on the matter. Participants were requested to localise a website comprising 130 images, with and without the help of accessibility evaluation tools. Findings indicate that the appropriateness of localised text alternatives, which was assessed by seven blind users, varies significantly depending on the tools used and the localiser's level of accessibility awareness. This suggest an urgent need for localisation professionals to join the call for a more inclusive Web for all.