X has begun a pilot that lets outside developers build “AI Note Writers” capable of drafting Community Notes, the crowd-sourced annotations the platform attaches to potentially misleading posts. The company said the program, launched on 1 July, is designed to increase the speed and scale of fact-checking while keeping “humans in charge”; notes generated by the bots will appear only if they receive high usefulness ratings from a diverse set of reviewers. Separately, Meta confirmed it is testing what leaked internal guidelines call “Project Omni”, an effort to train custom chatbots created in its no-code AI Studio to send users proactive follow-up messages. Under the rules, bots may contact a user only within 14 days of an initial conversation and after the person has sent at least five messages. Contractors at data-labeling firm Alignerr are helping refine chatbot personas so that replies reference earlier chats and avoid sensitive topics. The parallel initiatives underscore how social-media companies are embedding generative AI more deeply into their services to spur engagement and, ultimately, revenue. Court filings show Meta expects its generative-AI line-up to generate $2 billion to $3 billion in 2025 and projects as much as $1.4 trillion by 2035. X, meanwhile, is betting that automating the first draft of Community Notes can curb misinformation without sidelining human oversight.