Meta is testing a new AI feature that lets its chatbots send messages to users without being prompted. This marks a major shift in how artificial intelligence tools interact with users across Meta’s platforms.
The feature is part of a project called “Project Omni” and is currently being tested with custom AI bots through Meta AI Studio. These bots can remember past conversations, respond based on context, and now — for the first time — initiate chats themselves.
However, the bot won’t just message anyone. According to internal testing rules, it will only message users who have chatted with it at least five times in the past 14 days. The goal is to make interactions feel more natural, personalized, and engaging.
Meta hopes this leads to users spending more time on its apps, which could mean bigger earnings. Reports say the company could generate $2–3 billion in AI revenue this year, and that number might hit $1.4 trillion by 2035 if user engagement continues to rise.
A Meta spokesperson confirmed that bots will not randomly message users. Messages will only be sent if the user started a conversation first, and if ignored, the bot won’t try again. The move has raised privacy questions, but Meta insists the feature is designed to enhance user experience — not invade it.