Paper Type
Complete
Paper Number
PACIS2025-1910
Description
International interest in AI-driven UX/UI design for children’s digital platforms is growing. YouTube Kids, with over 145 million downloads globally, is among the most influential tools for children’s education. This study investigates how thumbnail elements affect children’s engagement with educational YouTube content. Using data from 38,598 videos collected in October 2024 from top-ranked children’s YouTube channels, we identify key visual factors influencing viewership. Results show that thumbnails with high color saturation, brightness, image quality, and the presence and sentiment of text increase engagement. Notably, thumbnails featuring objects rather than people drive more views, challenging conventional design strategies. Predictive modeling demonstrates significant improvement when thumbnail features are included, increasing R² from 0.9068 to 0.9253. These findings suggest that children’s media behavior differs from adults’ and requires tailored design approaches. This study offers insights for content creators and emphasizes the need for child-specific recommendation algorithms in educational platforms.
Recommended Citation
choi, sarah; Go, Young-Min; and Park, Keonchul, "Capturing Children's Attention: AI-driven Analysis of Thumbnail Design in YouTube Educational Content" (2025). PACIS 2025 Proceedings. 9.
https://aisel.aisnet.org/pacis2025/is_education/is_education/9
Capturing Children's Attention: AI-driven Analysis of Thumbnail Design in YouTube Educational Content
International interest in AI-driven UX/UI design for children’s digital platforms is growing. YouTube Kids, with over 145 million downloads globally, is among the most influential tools for children’s education. This study investigates how thumbnail elements affect children’s engagement with educational YouTube content. Using data from 38,598 videos collected in October 2024 from top-ranked children’s YouTube channels, we identify key visual factors influencing viewership. Results show that thumbnails with high color saturation, brightness, image quality, and the presence and sentiment of text increase engagement. Notably, thumbnails featuring objects rather than people drive more views, challenging conventional design strategies. Predictive modeling demonstrates significant improvement when thumbnail features are included, increasing R² from 0.9068 to 0.9253. These findings suggest that children’s media behavior differs from adults’ and requires tailored design approaches. This study offers insights for content creators and emphasizes the need for child-specific recommendation algorithms in educational platforms.
Comments
e-Learning