ChatGPT and the digital humanities: A cross-domain review of human-AI collaboration in education, work, and culture, and the question of trust for advancing SDG 4

Abstract

The rapid diffusion of ChatGPT into classrooms, workplaces, and caregiving settings has become a defining episode in the humanities' longstanding engagement with how new communication technologies reshape reading, writing, and human judgment — yet the scholarship on this episode remains fragmented, with most studies examining single settings in isolation rather than asking what these episodes share as instances of human-machine collaboration. This gap matters for the digital humanities in particular, where questions of language, literacy, authorship, and trust in mediated communication intersect with the broader cultural and ethical stakes of generative AI's integration into everyday human activity. This review addresses that gap by synthesizing evidence on ChatGPT usage, benefits, and trust concerns across workplace, educational, software development, marketing, and healthcare contexts, treating each as a site where humans and a language-generating machine negotiate authorship, judgment, and accountability. The purpose was to identify the task-level and domain-level conditions under which ChatGPT yields reliable gains, to characterize recurring reliability limitations, and to determine whether the structure of a task — rather than the sector in which it occurs — better explains when human trust in AI-generated language is warranted. A narrative multi-domain literature review was conducted, drawing on empirical studies published between 2023 and 2025 identified through structured database searches and reference-list screening, with findings extracted, categorized by domain, and synthesized thematically. Results showed that ChatGPT produced the most consistent gains on bounded, language-heavy tasks such as drafting, summarizing, and routine code generation, while performance declined on tasks requiring specialized judgment, cultural nuance, or verified accuracy, with reliability concerns — including fabricated citations and inconsistent outputs — recurring across otherwise unrelated sectors. These findings indicate that task structure, rather than industry sector, is the primary determinant of ChatGPT's trustworthiness, supporting adoption models that pair AI-generated language with calibrated human oversight in education, professional practice, and care — an approach consistent with the humanities' broader concern for preserving human judgment, authorship, and critical literacy amid the growing presence of generative language technologies in everyday life.