...

Making Authorship Visible: Assessing Generative AI in Multimodal Composing for Specific Purposes

Document Type : Review Articles

Author

School of Computing, Mathematics and Physics, University of Portsmouth, Portsmouth, United Kingdom

Abstract
Generative AI (GenAI) is capable of generating media content like voiceovers, scripts, images, and captions, which makes it impossible to determine learners’ competencies and contributions simply based on the existing multimodal products. This issue is highly relevant in the field of English and Specific Purposes Languages (ESP/LSP), as students design professional genres of communication intended for audiences that expect a high level of accuracy. This integrative-critical review explores the issue of GenAI-assisted Digital Multimodal Composing (RQ1), the challenges that come with assessing such products (RQ2), and potential principles of fair assessment of GenAI-created content (RQ3). The procedure included two rounds of online searching (39 queries in total), as well as backward citation tracing. Nineteen sources were referenced in this review, but out of 45 works read in full, only four were printed and read. The literature suggests that task design influences results, although the evidence for direct ESP/LSP outcomes still remains scarce. neither AI-use claims nor text-detection scores were capable of demonstrating any independent contribution. Recent frameworks are integrating process, product, and reflection, but there is still scarce evidence that shows whether AI use declarations are checked against something other than project results. As a solution, the author proposes the development of a provenance-based framework comparing the declaration against task-specific evidence, emphasizing the fact that corroboration cannot be considered a proof of authorship. The proposed framework has not yet been validated and does not claim any learning outcomes.

Keywords

Subjects


Articles in Press, Corrected Proof
Available Online from 15 December 2026