This study investigates the potential of prompt-based programming using generative AI, specifically large language models (LLMs), to support the development of computational thinking (CT) and programming skills in K-12 education. Through a series of workshops conducted with 9th-grade students in a Swedish public school, we compared two programming approaches: traditional block-based programming and prompt-based programming using a chatbot. Through a mixed-methods approach, we analyzed student performance across a range of algorithmic exercises, examined if the code delivered the requested output for each task, and assessed students’ self-reported engagement and learning outcomes. Results show that while block-based programming yielded higher success rates in the initial session, prompt-based programming surpassed BBP in later sessions in terms of success rate, particularly as task complexity increased and students became more familiar with prompt programming. Students reported higher engagement and perceived learning benefits when doing prompt-based programming. These findings suggest that prompt programming could have the potential of being a valuable complement to traditional programming methods, especially in supporting students with limited prior programming experience. However, further research is needed to explore how such tools can be effectively used in the classroom to foster genuine problem-solving skills and deep CT development.