Thread by @conspirator0

As Ukraine's presidential election approaches, we checked for signs of possibly related automated Twitter activity based on a review of tweets containing "Poroshenko" or "Zelensky" in Ukrainian, Russian, or English (220001 tweets / 33420 accounts over 10 days.)

cc: @ZellaQuixote

(search used to obtain this dataset)

lang:en OR lang:uk OR lang:ru Poroshenko OR Пороше́нко OR Порошенко OR Zelensky OR Зеленский OR Зеленський

4.3% of the tweets about the candidates in Ukrainian are posted via apps/automation services. This percentage more than doubles for Russian (10.3%) and English (9.5%). Can we rule out this anomaly being an across-the-board difference in automation levels between languages?

We looked at 24 hours' worth of tweets from four different "neutral" searches in Ukrainian, Russian, and English, and the automation percentages showed minimal variance between languages. Only the tweets about the candidates show a significant difference between languages.

The bots involved look like many smaller networks rather than one large one. Some botnets are more competent than others, and this here group of six accounts can't manage the simple task of copying Russia Today tweets without dropping the link to the actual article. #YouHadOneJob

Another set of copycat bots (seven this time), with a twist: if the tweet they're cloning isn't in Russian, they translate it first.

One more group of bots for the road: 12 following the naming convention [x]news777 or [x]news888 (example, @crimeanews777). These all spam links to hypothetical "news" sites matching the account name.