A new study has revealed that up to 15 percent of Twitter accounts are not run by humans but active bots controlled by software.

The microblogging site boasts of about 319 active users per month. This translates to about 48 million bot accounts, according to a research team from the University of Southern California (USC).

Bots are automated software used mainly to collect data.

The figure is less than the 8.5 percent estimate of bot accounts filed by Twitter with the SEC.

According to USC, the bots can 'like,' 'retweet,' and 'follow' like a normal human does. 

Moreover, they can also interact socially by disseminating news and publications, as well as coordinate volunteer activities.

Unfortunately, they, too, can support malware.

To determine which accounts are legit or fake, the researchers used some features on Twitter such as friends, activity time series, tweet content and sentiment, and network patterns.

Meanwhile, the BBC also reported earlier this year, that UK researchers from University College London (UCL) also studied bots on Twitter.

Some of the accounts were found to having been used to send spam, boost traffics in trending topics, and to fake follower numbers.

The research team led by a graduate student of computer science Juan Echeveria started by randomly selecting a sample of 1 percent of Twitter users to gauge how people utilize such social network.

As the analysis of the data progressed, however, Echeveria observed some strange results that point to lots of linked accounts being run by a botnet and operated by a single person or group.

These bots stood out as they all shared distinct features such as messages being posted only from Windows phones, tweets from places where nobody lives, and almost exclusively include quotes from Star Wars novels.