Regardless of the content of tweets, it’s the timing of those tweets that can help distinguish bots on Twitter from actual individual and corporate accounts.
Aldo Faisal and Gabriela Tavares, two researchers from Imperial College London, studied more than 160,000 tweets from personal Twitter accounts, company (“managed”) accounts and accounts controlled by bots (found by using online lists of Twitter bots). Their research was published last week. An excerpt from the summary:
Periods of high or low Twitter activity and the time between successive tweets could distinguish the three kinds of accounts from one another with approximately 83% accuracy.
The study found that personal accounts are most active in the afternoons and evenings (based on local time of the accounts studied), and — not surprisingly — corporate accounts are most active during business hours. The tweet distribution for bot accounts, however, “exhibits a variety of peaks, which is probably because their behaviour is not associated with a structured daily routine.” That’s shown in the chart below, which shows tweet activity throughout a 24-hour period.
No doubt the spam team at Twitter looks at timing in its anti-bot efforts, so this isn’t likely to be surprising to anyone at Twitter HQ.
And I should mention, too, that not all Twitter bot accounts are spam. There are countless automated accounts that push out things like news headlines, weather updates, sports scores and more.