In the present world, where educational and professional achievement involves openness and the capacity to assert oneself, this attribute finds its place in interview strategies for efficient selection. This paper describes a multimodal approach to analyzing and estimating confidence levels using an AI-based system for interpreting important information from the primary source, such as faces and voices. One unique aspect of this system is the ability to distinguish fake confidence. This is done by adopting time-based questioning, speech-text relevance, question correctness analysis, and video and audio analysis employing a proprietary algorithm. Combining these modalities, this system was designed to give a holistic and likely accurate depiction of an individual’s confidence during different tasks, with a specific focus deemed important for educational and/or interview contexts. In conclusion, this system offers a very strong basis and support for estimating confidence and identifying fake confidence, making it a system that fosters a culture of open and constant improvement. This system shows a promising RMSE of about 9.32 and an average F1-score of 0.82. This indicates that the scale effectively identifies the true confidence level per those results.