ORIGINAL RESEARCH article
Front. Artif. Intell.
Sec. AI for Human Learning and Behavior Change
Volume 8 - 2025 | doi: 10.3389/frai.2025.1690616
This article is part of the Research TopicNew Trends in AI-Generated Media and SecurityView all 8 articles
Explainable Multilingual and Multimodal Fake News Detection: Towards Robust and Trustworthy AI for Combating Misinformation
Provisionally accepted
- 1College of Engineering, Bharati Vidyapeeth (Deemed to be University), Pune, India
- 2Vishwakarma Institute of Technology, Pune, India
- 3MIT Art Design and Technology University, Pune, India
- 4Vishwakarma University, Pune, India
- 5San Jose State University, San Jose, United States
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Abstract. Fake news detection requires systems that are multilingual, multimodal, and explainable—yet most existing models are English-centric, text-only, and opaque. This study introduces two key innovations: (i) a new multilingual–multimodal dataset of 74,000 news articles in Hindi, Gujarati, Marathi, Telugu, and English with paired images, and (ii) HEMT-Fake, a Hybrid Explainable Multimodal Transformer that integrates text, image, and relational signals with hierarchical explainability. The architecture combines transformer embeddings, CNN–BiLSTM text encoders, ResNet image features, and GraphSAGE metadata fused via multi-head attention. Its explainability module unites attention, SHAP, and LIME to provide token-, sentence-, and modality-level transparency. Across four languages, HEMT-Fake delivers a ~5% Macro-F1 improvement over XLM-R and mBERT, with gains of 7–8% in low-resource languages. The model sustains 85% accuracy under adversarial paraphrasing and 80% on AI-generated fake news, halving robustness losses compared to baselines. Human evaluation shows 82% of explanations judged meaningful, confirming transparency and trust for fact-checkers. Impact Statement.HEMT-Fake advances fake news detection by combining multilingual coverage, multimodal reasoning, and explainable outputs in a single framework. By achieving higher accuracy in low-resource languages and maintaining robustness against AI-generated misinformation, it directly supports fact-checkers, journalists, and policymakers in combating misinformation across diverse linguistic and cultural contexts.
Keywords: Fake news detection, Misinformation and disinformation, Multilingual Dataset, Explainable artificial intelligence, Hybrid Deep Learning Architecture, Adversarial robustness, social media analysis
Received: 22 Aug 2025; Accepted: 22 Oct 2025.
Copyright: © 2025 Jadhav, Meshram, Bhosle, Patil, Dash and Jadhav. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence:
Kailas Patil, kailas.patil@vupune.ac.in
Shrikant Jadhav, shrikant.jadhav@sjsu.edu
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.