This study explores how different data representations support reflection in personal informatics, specifically in the context of fitness tracker data. The research compares text generated by a Large Language Model (LLM), standard charts, and a combination of both. The study involved 273 participants and used a between-subjects design with three conditions: text-only, chart-only, and a combination of both. The findings indicate that users experienced more reflection, focused attention, and reward when presented with LLM-generated qualitative data compared to standard charts alone. The study demonstrates that automatically generated data descriptions can effectively complement numeric fitness data, fostering a richer and more reflective engagement with personal wellbeing information. The research contributes to the HCI community by providing insights into how AI can be used to enhance reflection in personal informatics systems.This study explores how different data representations support reflection in personal informatics, specifically in the context of fitness tracker data. The research compares text generated by a Large Language Model (LLM), standard charts, and a combination of both. The study involved 273 participants and used a between-subjects design with three conditions: text-only, chart-only, and a combination of both. The findings indicate that users experienced more reflection, focused attention, and reward when presented with LLM-generated qualitative data compared to standard charts alone. The study demonstrates that automatically generated data descriptions can effectively complement numeric fitness data, fostering a richer and more reflective engagement with personal wellbeing information. The research contributes to the HCI community by providing insights into how AI can be used to enhance reflection in personal informatics systems.