The Internet Dies When Bots Write for Bots to Read
In September 2018, a Proceedings of the National Academy of Sciences study analyzed more than 13 million Twitter posts containing links. The researchers found that a small group of highly active accounts helped low-credibility content spread during its early stage. Many had bot-like features: they were software able to post, reply, share, follow, or collect data. They did not need to persuade readers with an argument. They only needed to make a link appear often enough and early enough for others to assume that a crowd was interested.
⬛ CONFIDENTIAL nhat141210/data 8/15/2026 1 min read
In September 2018, a Proceedings of the National Academy of Sciences study analyzed more than 13 million Twitter posts containing links. The researchers found that a small group of highly active accounts helped low-credibility content spread during its early stage. Many had bot-like features: they were software able to post, reply, share, follow, or collect data. They did not need to persuade readers with an argument. They only needed to make a link appear often enough and early enough for others to assume that a crowd was interested.
In September 2018, a Proceedings of the National Academy of Sciences study analyzed more than 13 million Twitter posts containing links. The researchers found that a small group of highly active accounts helped low-credibility content spread during its early stage. Many had bot-like features: they were software able to post, reply, share, follow, or collect data. They did not need to persuade readers with an argument. They only needed to make a link appear often enough and early enough for others to assume that a crowd was interested.
In September 2018, a Proceedings of the National Academy of Sciences study analyzed more than 13 million Twitter posts containing links. The researchers found that a small group of highly active accounts helped low-credibility content spread during its early stage. Many had bot-like features: they were software able to post, reply, share, follow, or collect data. They did not need to persuade readers with an argument. They only needed to make a link appear often enough and early enough for others to assume that a crowd was interested.
In September 2018, a Proceedings of the National Academy of Sciences study analyzed more than 13 million Twitter posts containing links. The researchers found that a small group of highly active accounts helped low-credibility content spread during its early stage. Many had bot-like features: they were software able to post, reply, share, follow, or collect data. They did not need to persuade readers with an argument. They only needed to make a link appear often enough and early enough for others to assume that a crowd was interested.