Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{210780,
        author = {Rahuf M and Dr. Bonthu Kotaiah},
        title = {AI Based Detection of Misleading Content by Comparing Images with Captions},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {6},
        pages = {2109-2119},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=210780},
        abstract = {With the prevalence of image sharing on social media, a growing concern is the possibility that an image and its caption may present misleading or even contradictory statements about what is actually shown in the image. The caption could misrepresent the people, location, timing, or events shown in the image. The usual approaches to detecting fake news from text are not sufficient for addressing this issue since they focus only on the caption text, not how it relates to the accompanying image. This paper proposes an AI-based approach to identifying potentially misleading image-caption pairs. The current prototype focuses on a vision-language model to analyze an image-caption pair, provide a risk score, identify any evidence mismatch, and generate an explanation. The prototype also includes a heuristic fallback mode, a web interface, and database storage of results. The analytical framework focuses on object, entity, event, temporal, and contextual consistency with the image content, and attempts to classify evidence as supported, contradicted, or unverifiable. The final classification output has three possible labels: consistent, misleading, and uncertain. Unlike previous work on the fake news detection task, this proposed framework is focused on detecting unsupported contextual claims and provides an explicit consideration of evidence mismatch and uncertainty. This prototype is not a fully formed fact-checking system, but rather a research tool that will be evaluated in practice on a future annotated image- caption dataset to formally establish its classification accuracy, explanation quality, robustness, and generalizability.},
        keywords = {Multimodal misinformation detection, image-caption consistency, vision-language models, contextual misinformation, artificial intelligence, explainable AI, cross-modal verification.},
        month = {November},
        }