Study Finds 734,000 Chinese Bot Accounts Flooding X's Japanese Feeds

A team of Japanese and Chinese researchers spent nearly two years tracking something most of us never see: a massive, automated spam operation running inside X's Japanese-language feed. They identified 734,173 accounts posting in Chinese that behaved in lockstep — copying the same text over and over — and collectively published about 495 million posts between May 2024 and March 2026. Applying the same test to Japanese-language accounts in the same period caught six times fewer suspicious accounts, meaning this flood was unusually concentrated.

The surprise is what the bots were actually saying. Most of the content was commercial, not political — think repetitive advertising, link spam, and self-promotion rather than election meddling. The researchers describe the activity as "self-referential," meaning the accounts largely engage with each other, creating the illusion of popularity without needing real people. It's the online equivalent of a crowd of people clapping for each other in an empty room.

Where real users did show up matters. When the researchers looked at posts flagged as potentially harmful or illegal, they found a comparatively high share of reactions coming from outside the Chinese-speaking population — in other words, ordinary Japanese users were the ones engaging with the genuinely nasty stuff. That's a useful reminder that fake accounts don't need to fool everyone to do damage; they just need to push certain content in front of real eyes.

The practical takeaway for you: follower counts, reposts, and trending topics are softer signals than they look. If a post feels like it was written by a copy-paste machine, it probably was. The study also gives platforms a clearer fingerprint for spotting this kind of noise — repeated text, coordinated timing, and a closed loop of self-promotion — which could eventually mean cleaner feeds for everyone.