A study by the Oxford Internet Institute and the Hasso Plattner Institute has highlighted how AI-mediated communication can alter the tone of social media posts, potentially shifting public opinion over time. AI tools can rewrite posts with a consistent directional bias, observed across different AI models, nudging posts on topics like feminism and gun control in specific directions. While a single AI-edited post may seem insignificant, the cumulative effect of millions of similar edits can influence community opinions over time. AI tools can also alter a post's tone, making it sound more sympathetic or critical, and can implicitly frame debates by emphasizing certain aspects over others. Bias can be introduced through underlying instructions, such as platform-level design choices seen in an audit of Grok’s ‘Explain this post’ feature. AI tools used to moderate content can also be biased, affecting which opinions gain traction. The instructions guiding AI tools are often hidden from users, making their influence hard to detect.