An Adaptive Intent-Centric AI Framework for Content Personalization on Intelligent Digital Platforms

Abstract

As digital platforms have grown rapidly, there is a growing need for intelligent content personalisation systems that can provide real-time, relevant, and engaging recommendations. In contrast, traditional recommendation methods are not ideal for capturing dynamically changing user preferences, contextual behaviour, and recommendation transparency, all of which can decrease user satisfaction and engagement. In response, this paper presents an Adaptive Intent-Centric Personalisation Framework (AICP-Former) for AI-based content recommendation. It proposes an Adaptive Intent-Centric Personalisation Framework (AICP-Former) for AI-based content recommendation. The proposed framework consists of a Context-Aware Preference Memory Network to store short-term and long-term behavioural patterns, an Adaptive User Intent Graph to model changes in user preferences, and a Multi-Agent Reinforcement Personalisation Engine to optimise recommendation relevance, diversity, freshness, and engagement collectively. A Trust-Aware Explainability Module offers interpretable recommendation decisions, and a Dynamic Feedback Optimisation mechanism optimises the user profiles in real time based on feedback and interactions. The experimental evaluation shows that the proposed framework can reach 97.36% recommendation accuracy, 96.92% precision, 96.58% recall, and 96.75% F1. Furthermore, the framework provides a high processing throughput of 1688 requests/s with only 68 ms recommendation latency; the click-through rate and engagement score increase to 16.2% and 95.3%, respectively. The findings validate the effectiveness of the proposed AICP-Former in improving recommendation quality, user engagement, computational efficiency, and scalability, making it a potential solution for next-generation intelligent digital platforms.