Gender Bias and Literary Meaning in AI-Mediated Translation: A Critical Review of Multilingual Evidence

Abstract

Artificial intelligence has expanded the speed and accessibility of multilingual translation, yet its consequences for gender representation and literary meaning require more precise claims than are often made. This critical narrative review synthesises evidence from ten recent publications in Scopus-indexed journals or conference proceedings, three directly relevant articles from Acta Humanitatis, DIALOGICA, and AI, Law, Politics, and selected foundational works in discourse, gender, and translation studies. It asks which forms of gender bias are documented in machine translation and large language models, how those findings intersect with the demands of literary translation, and what research protocol is needed for accountable human–AI interpretation. The literature provides convergent evidence of masculine defaults, occupational stereotyping, and failures to preserve gender-neutral or ambiguous forms in defined multilingual settings. It also shows that surface fluency does not guarantee creativity, narrative engagement, poetic force, or intercultural adequacy. At the same time, benchmark findings cannot be generalised automatically to every language pair, model, prompt, genre, or literary corpus. The review therefore rejects both technological determinism and uncritical claims of algorithmic agency. It proposes a transparent protocol centred on identifiable texts, model and prompt documentation, repeated runs, preserved outputs, bilingual evaluation, linguistic evidence, negative cases, and explicit human responsibility. An intergenerational research agenda is outlined, while no age-related empirical effect is claimed.