Generative AI and similar emerging technologies are increasingly being discussed in the context of the processing of human language (Leivada et al., 2022). DALL·E 2 and other AI image generators are capable of producing unique images from text prompts using a huge catalogue of online source data. Multiple semantic elements contained in the lexical and syntactic components of texts prompts can be recreated in visual format signifying a large step forward in generative AI technology. As tools such as DALL·E 2 develop, their potential applications in language teaching and assessment will become more relevant. Major testing institutions are working on incorporating AI into testing and educational materials, and it is thus imperative to maintain a clear understanding of the capabilities of such tools.
While previous studies have investigated the relationship between prompt and output in DALL·E 2 (Conwell & Ulman, 2022; Leivada et al., 2022; Marcus et al., 2022; Thrush et al., 2022), this may be the first study with an explicit focus on the applications of generative image AI tools in language teaching and assessment. The three strands of this paper approach the analysis of the tool from different perspectives using a range of methods. The aim is to provide a broad overview of its capabilities for the purposes of language assessment and discuss the role of generative AI technologies in the field.