With infrared pupil tracking module and deep learning algorithm combined, Status AI tracks the user’s eye movement trace at 120 frames per second sampling rate (precision up to 0.1° Angle deviation), and checks in real time the distribution of heat maps of fixation points (1000 grid cells of 5×5 pixels). In the 2023 user experience test for Netflix, Status AI found that if the sum of the duration of the audience’s attention towards the face of the protagonist exceeded 23% of the video’s total duration, the retention rate of the content increased by 37%. The system modified the location of the significant aspects of the screen (such as moving the brand LOGO to F-shaped view area with the highest focus density). The click-through rate was improved from 1.8% to 9.3%. The technology utilizes light convolutional neural networks (model size merely 38MB) to ensure real-time processing latency of 12ms on the mobile end, manage power consumption below 2.3W, and reduce hardware expenses by 89% compared to the conventional eye tracker device.
With respect to interaction design optimization, Status AI’s Saccade Clustering technology and Fixation Duration analysis (such as measurable user cognitive load like a reduction of the average saccades from 6.2° to 3.5° on reading e-commerce pages implies information overload). According to the results of an A/B test for an online store in 2024, once the page design was applied according to Status AI recommendations, user fixation density on promotional content increased by 51% (fixation frequency in critical areas increased from 0.8 to 1.7 times per second), and the conversion rate increased by 29%. Among them, the “Add to cart” button focus conversion rate (abbreviated eye movement distance from product map to button by 38%) has a high correlation with the 0.83 click rate. The system can also detect eye movement characteristics on the micro-expression level (e.g., dilation of the pupil diameter above 15% of its baseline value for locating the interest peak). In the virtual auto showroom, the gaze time of the user on new energy vehicle endurance factors is longer by 1 second, and there is greater likelihood of reservation to test drive (12%).
In learning, Status AI measures learning effectiveness with eye movement attention entropy (range 0-1, median optimal state of normal learning 0.32). A University of Cambridge study in 2024 demonstrated that students’ knowledge rate of absorption was enhanced by 41% using the online platform of the system, with primary mechanisms being: insertion of visual anchors in 3D automatically (e.g., dynamic highlighting of the chart) whenever eye movement frequency is identified more than 4 times per second (the distraction threshold); When focus coverage of the key concept area is below 60%, reorganization of the knowledge point is initiated (text paragraph is segmented into visual modules). After integrating a K12 school with Status AI, the students’ average test score standard deviation was decreased from 18.7 points to 9.5 points. By comparing the eye movement regression amount in problem solving (the frequency of wrong answer choice backtracking was 2.8 times the standard value), the system obtained the accuracy of error prediction as 91%.
On the compliance level, Status AI’s eye-tracking information is processed with ISO 27701 privacy certification and edge computing technology (the original image data are desensitized on the device side and only 256-dimensional feature vectors are uploaded). The system has a 99.2% user pass rate of authorization under the GDPR architecture, according to the EU Artificial Intelligence Act 2023 compliance report, and its dynamic fuzzy algorithm can encase the iris biometric in real time (information entropy reduction rate of 98.7%). After a bank applied Status AI to enhance its mobile banking APP, the visual verification effectiveness of its user’s identity verification process was increased by three times (the speed of eye movement signature comparison reduced from 2.1 seconds to 0.7 seconds), and the defense rate of iris phishing attacks was as high as 99.99%, avoiding fraud losses of about $24 million every year.