Table 9.

Contributions.

Dimension	Previous studies	Contributions of the present study	Added value compared to previous research
Scope and focus	Most studies are either general reviews (Mujtaba et al., 2017; Qabajeh et al., 2018; Kunju et al., 2019; Athulya and Praveen, 2020) or broad bibliometric analyses on malware, phishing in general, or phishing and Big Data (Mat et al., 2021; Mutluturk and Metin, 2023; Pejić-Bach et al., 2023)	Bibliometric analysis focused exclusively on AI for phishing detection, complemented by a content analysis of the 10 most cited papers in the dataset	ML and DL-centric overview of phishing detection, filling a gap in previous reviews
Temporal coverage	2021–2025	Updates the temporal landscape by showing significant growth: 2024 is the most productive year (≈2 × 2023), with 2025 continuing the upward trend	Up-to-date perspective
Technological transition to DL	DL is mentioned as promising	Documents the transition from classical ML to DL and hybrid/stacking models, explaining the drivers: scalability, zero-day detection, FP reduction, and improved accuracy	Updated technological evolution
Practical implications	Mostly focused on academic models	Integrates practical examples: Microsoft Defender, Google Workspace, Barracuda, TitanHQ, Trustifi, Abnormal Security	Presentation of the implementation layer: AI-powered, real-time, behavioral, NLP, and computer vision-based detection of phishing
Integration into enterprise security	Not identified	Proposes integration of AI-powered phishing detection into SIEM, endpoint protection, secure email gateways, and cloud-based defense systems	Bridges between theoretical research and applied cybersecurity