•  
  •  
 

Journal of Arabic Sciences and Humanities

Corresponding Author

Fahad Ahmed Otaif

Authors ORCID

Fahad Ahmed Otaif: https://orcid.org/0000-0001-9832-4289

Article Type

Article

Keywords English

Artificial intelligence, Written compositions, Discourse processes, AI-authored discourse, Narrative

Abstract English

With the growing use of generative AI bots in different disciplines and professions, AI bots have been increasingly adopted to compose, revise, translate, and author content, yet without much research evaluating their quality, particularly when it comes to the processes and features of AI-authored discourse. The current study investigates, based on corpus-based evidence, the extent to which one-theme AI-generated sentences can be reorganised/turned into meaningful discourse through four AI bots (ChatGPT, Gemini, Copilot, and DeepSeek), i.e. by which discourse processes and in what coherent and cohesive patterns. These four AI-bots were fed with well-designed prompts instructing them to produce both fictional and non-fictional discourse on a single topic, which is coffee. The resulting texts formed a corpus of AI-generated discourse, which was then analysed using a qualitative discourse analysis approach. AI bots were found to construct their discourse through several processes, including 1) text segmentation, 2) sentence reorganisation based on genre-relevant themes, 3) lexical and information omission to maintain cohesion and coherence, and 4) paraphrasing, which includes creative additions or simplifying original sentences to match the target audience's level. AI bots generally processed and prioritised information according to relevance, audience, and order of appearance in the prompt, although some outputs failed to simplify their discourse appropriately for the intended audience. These findings show how corpus-based discourse analysis can clarify both the strengths and limitations of AI-authored texts and help refine our understanding and use of this rapidly expanding authoring technology. Future research should explore how AI bots handle multimodal discourses, including emojis.

Digital Object Identifier (DOI)

10.71422/1658-4058.1014

Accept Date

28 February 2026

Publication Date

4-25-2026

Download

Share

COinS