Troll farm

A troll farm is staffed by people, not software. Employees work in shifts, follow instructions on themes and targets, run personas with invented identities and post in different tones for different audiences. Accounts may be built over months to look like ordinary members of a community before they push a message. Operators sometimes add bot accounts for scale, but the human operation is what separates a farm from an automated network.

Customers vary. Some farms work for governments, some for political campaigns and some are commercial services hired by whoever pays. The motive can be persuasion, distraction, division or harassment of a person or company.

The best-documented case is the Internet Research Agency in St. Petersburg. A 2018 US Department of Justice indictment described how its staff used fake American personas on social platforms from 2014 onward, and platforms later removed accounts linked to it. Researchers working with the US Senate studied its output in detail. The case is useful because the primary documents are public.

A farm can be pointed at a company as well as at an election. A hypothetical example: a coordinated set of accounts amplifies a false claim about a product during a launch, and journalists begin quoting the volume of posts as evidence of public anger. Analysts look at account creation dates, posting rhythm, shared content and the network that links accounts before concluding a surge is organic. Platforms use the term coordinated inauthentic behavior for this activity, and the broader field is information operations. Related tactics include astroturfing and sock puppets.

An angry crowd is not necessarily a troll farm. Organic outrage and brigading by real people look similar from outside.

Farms are not always foreign or state-run, and a farm's presence in a conversation does not show it created the anger or persuaded anyone. Naming an operator requires evidence beyond account behavior.