September 24, 2026
Bots and coordinated accounts on social media: detecting them does not prove they had influence
A study proposes measuring how much spread is really due to coordinated accounts. In the 2019 UK election, very little. In other campaigns, as much as the most central accounts.
Every time a controversy grows on social media the same explanation appears: "it was bots". Sometimes there really were coordinated accounts. But their presence does not mean they were the ones that moved the conversation. A study published as a preprint on 21 September by Niccolò Di Marco, Matteo Cinelli, Shinichi Nakano and Andrea Frosini focuses on that second step: measuring influence after detection.
What they propose
Two complementary methods. The first reconstructs the diffusion cascades (who shares from whom) and calculates how much influence the coordinated accounts had against the maximum they could in theory have reached. The second works on the repost network and compares the coordinated accounts with the best-placed accounts in that network.
What they find
- 2019 UK general election: coordinated accounts achieved only a small fraction of the influence they could in theory have reached.
- State-linked information operations in several countries: large differences between campaigns. In several, the coordinated accounts held advantageous positions in the network and their influence was comparable to, or even greater than, that of the most central accounts.
The two methods agree, which suggests the result depends on the structure of each network and not on how it is measured.
Why it matters
Because answering "were there bots?" almost never closes the case. For a public body, an institution or a party, reacting to a campaign that moved nothing can give it the visibility it lacked. And not reacting to one that did carry weight is the opposite mistake.
How we read it at Trendia
In an analysis of public opinion on social media it helps to separate four questions that tend to get mixed up:
- Was there coordination?
- Where were those accounts in the conversation, at the centre or on the edge?
- Would the content have circulated anyway without them?
- How much of the spread can be attributed to them?
The third is the one most often forgotten, and it means looking at the content as well as the accounts: what was said, in what format and with what emotion. A narrative that already resonated with people does not need much help to grow.
The study does not offer a universal figure. Its contribution is a more demanding way of asking the question.
If you work in a public body, an institution or a party, here is how we analyse public opinion on social media.
Source: Optimal and heuristic strategies for evaluating the influence of coordinated behavior in information cascades and retweet networks, arXiv, 21 September 2026.