Generative artificial intelligence (GenAI) is reshaping media and communication, yet research on its semiotic role in meaning-making remains theoretically and methodologically fragmented. This systematic review examines semiotic frameworks, technological modalities, communication contexts, and human–AI agency in research published from 2021 to 26 August 2026. Scopus and Web of Science searches identified 376 records; the final included corpus comprises 128 studies. The synthesis distinguishes between modality-specific concerns: studies of large language models primarily examine interaction and dialogue, whereas text-to-image studies emphasize representation and synthetic imagery. The RAW agency data contained four cases coded as predominantly AI-led, but these classifications changed following a documented text-consistency review, showing that the agency pattern is sensitive to coding decisions. We derive four provisional configurations: instrumental human anchoring, mediated distributed semiosis, co-creative/agentive participation, and contested or unresolved computational semiosis. The studies also reveal interconnected methodological, conceptual, cultural, technological, and governance gaps. These findings suggest that participation in sign production and mediation should be distinguished from claims of autonomous intentional agency; they do not establish whether generative systems possess intentional understanding.

Academic engagement and perceived employability are increasingly relevant to how higher education prepares students for changing labor markets, yet their empirical relationship remains theoretically and methodologically fragmented. This systematic review synthesized primary studies that substantively examined both constructs among higher education students. Reporting followed PRISMA 2020. The original Scopus and Web of Science searches were expanded during peer review to ERIC, APA PsycINFO, Education Source, and ProQuest, and one-generation backward and forward citation searching was completed. Across all sources, 627 raw records were consolidated, 119 duplicates were removed, 508 unique records were screened, 17 full texts were assessed, and 14 studies met the eligibility criteria. Evidence was synthesized narratively because study designs, measures, modeled directions, and statistical metrics were not sufficiently comparable for meta-analysis. Twelve studies were cross-sectional and two were longitudinal or time-lagged. The evidence was predominantly consistent with a positive association between engagement and perceived employability, but effect magnitude, dimensional patterns, modeled direction, and explanatory mechanisms varied. Engagement frequently operated as a mediator linking educational, relational, or psychological resources with employability perceptions, whereas several studies modeled perceived employability as an antecedent of engagement. JBI appraisal indicated generally adequate measurement and statistical analysis but recurrent limitations in explicit confounding control and temporal inference. The findings support a cautious, mechanism-oriented interpretation: academic engagement and perceived employability are related, but causal ordering remains unresolved. Future research should prioritize longitudinal, reciprocal, dimension-sensitive, multi-source, and cross-context designs and directly measure sustainable career outcomes rather than treating perceived employability as a proxy for career sustainability.

Artificial intelligence (AI) is transforming the accounting function, but evidence from emerging economies still leaves a central question unresolved: whether the profession’s traditional strengths are sufficient to support this technological transition. This study examines this issue in Peru, where digital reform has advanced rapidly, but professional adaptation remains uneven. Using survey data from 424 accounting professionals and academics, a Covariance-Based Structural Equation Modeling (CB-SEM) approach was employed to test the relationships between Sustainability, Professional Challenges, IFAC Ethics, and AI Orientation. The construct previously described as AI adaptation is now defined more cautiously as AI Orientation, because its indicators capture a broad perceptual orientation toward AI, including perceived challenge, workflow impact, curricular integration, substitution concerns, and routine use, rather than verified adoption behavior. The results confirmed the strength of the traditional professional sequence: Sustainability significantly predicted Professional Challenges (β = 0.637; p < 0.001), and Professional Challenges strongly predicted adherence to IFAC Ethics (β = 0.972; p < 0.001). However, the direct path from Sustainability to IFAC Ethics was not significant (β = 0.200; p = 0.096). Regarding technological integration, the model revealed a structural disconnect. While Sustainability showed a weak direct effect on AI Orientation (β = 0.279; p = 0.045), neither Professional Challenges (β = 0.160; p = 0.510) nor IFAC Ethics (β = 0.037; p = 0.810) significantly predicted AI Orientation. These findings indicate that while the ethical and adaptive core of the Peruvian accounting profession remains highly resilient under pressure, its connection to AI is still fragmented, with algorithmic tools often perceived as external technological systems rather than integrated accounting competencies.