As more research papers are read on tablets and e-readers, readers struggle with dense texts with technical terms and novel concepts. In response, generative AI tools are being developed to support faster and more effective comprehension. A previous study introduced a Web-based interface that uses ChatGPT to display key information arranged around a research paper PDF, but it lacked the ability to respond dynamically to the reader’s current focus within the document. This paper presents an iPad application that identifies the sentence being read using a digital reading ruler and provides in-context paraphrasing and reflective summaries to support comprehension of research papers. The system provides multilingual support powered by GPT-4o, which readers can flexibly access as needed. In our user study, participants read both Japanese and English papers using three media—our application, the default iPad PDF viewer, and printed paper–and evaluated them through comprehension tests and usability questionnaires. Results showed that the implemented support features were rated as useful by at least half of the participants, with many also reporting improved subjective reading speed and depth of understanding.