Speakers of different languages (e.g., English vs. Turkish) show a binary split in how they package and order components of a motion event in speech and co-speech gesture, but not in silent gesture (Özçalışkan et al., 2016b). In this study, we focused on Chinese that does not follow the binary split in its expression of motion in speech (Slobin, 2004), and asked whether adult Chinese speakers follow the language-specific speech patterns in co-speech but not silent gesture, thus showing a pattern akin to Turkish and English adult speakers in their description of animated motion events. Our results provided evidence for this pattern, with Chinese—as well as English and Turkish— speakers following language-specific patterns in speech and co-speech gesture, but not in silent gesture. Our results provide support for the “thinking-for-speaking hypothesis” (Slobin, 1996), namely that language influences thought only during online, but not offline, production of speech.