Gendered style transfer aims to rewrite text toward linguistic styles commonly associated with feminine or masculine language use while preserving the original meaning. Prior work commonly operationalizes the transferred target style through binary author identity labels, conflating who writes a text with how gender expression is conveyed stylistically in the text itself. In contrast, we focus on perceived gendered style: the qualities attributed to a text by readers, independently of the author’s identity. We study this as a controlled contrastive rewriting problem: given the same source text, can different prompt-based operationalizations induce feminine- and masculine-targeted rewrites, and at what semantic cost? We compare four operationalizations across instruction-tuned LLMs: author identity, gendered voice, general style descriptions, and fine-grained sociolinguistically grounded style instructions, and evaluate them against human-written rewrites. Using automatic metrics, statistical tests, and human evaluation, we find that LLM-generated rewrites often show stronger contrast and higher meaning preservation than human-written ones. Moreover, prompt operationalization systematically shapes the trade-off between perceived stylistic contrast and semantic fidelity. These findings provide practical guidance for prompt design in gendered style transfer and suggest that targeting perceived style with LLMs can yield clearer linguistic signals than relying solely on author-identity-based human rewrites.

In NLP, language–gender associations are commonly grounded in the author’s gender identity, inferred from their language use. However, this identity-based framing risks reinforcing stereotypes and marginalizing individuals who do not conform to normative language–gender associations. To address this, we operationalize the language–gender association as a perceived gender expression of language, focusing on how such expression is externally interpreted by humans, independent of the author’s gender identity. We present the first dataset of itskind: 5,100 human annotations of perceived gendered style—human-written texts rated on a five-point scale from very feminine to verymasculine. While perception is inherently subjective, our analysis identifies textual features associated with higher agreement among annotators: formal expressions and lower emotional intensity. Moreover, annotator demographics influence their perception: women annotators are more likely to label texts as feminine, and men and non-binary annotators as masculine. Finally, feature analysis reveals that the text’s perceived gendered style is shaped by both affective and function words, partially overlapping with known patterns of language variation across gender identities. Our findings lay the groundwork for operationalizing gendered style through human annotation, while also highlighting annotators’ subjective judgments as meaningful signals to understand perception-based concepts.

Authorship Profiling (AP) aims to predict the demographic attributes (such as gender and age) of authors based on their writing styles. Ever-improving models mean that this task is gaining interest and application possibilities. However, with greater use also comes the risk that authors are misclassified more frequently, and it remains unclear to what extent the better models can capture the bias and who is affected by the models’ mistakes. In this paper, we investigate three established datasets for AP as well as classical and neural classifiers for this task. Our analyses show that it is often possible to predict the demographic information of the authors based on textual features. However, some features learned by the models are specific to datasets. Moreover, models are prone to errors based on stereotypes associated with topical bias.