Dawn Chmielewski and Katie Paul by Sheila Dang

(Reuters) – Twitter (NYSE.:) has reduced the use of fake accounts since 2013. Independent researchers have stated that less than 5% are spam or false accounts.

The discrepancy may now impact the outcome of Elon Musk’s $44billion cash deal. Musk, a billionaire, tweeted that Friday’s takeover bid was “temporarily suspended” as he searched for information on fake Twitter accounts.

Musk may have used this tactic to negotiate, but the fact is, almost nothing is known about how accounts are defined, or disregarded, according to independent social media research and current employees.

According to a public filing, the social media platform stated in May 2, that less than 5% of its 229,000,000 daily active users are being targeted by advertising. This was based upon an internal analysis of some of its accounts. This figure does not account for pseudonymous, automated and parody profiles allowed on the social media platform.

Twitter didn’t respond to our requests for comment.

Based on a 2017 study and recent research by a company that monitors internet conversations, researchers estimate that between 9% and 15% of all the Twitter accounts are bots.

Dan Brahmy (CEO of Cyabra, an Israeli tech company using machine learning for fake accounts identification) said “They’ve underestimated that number.”

Cyabra estimated that 13.7% of Twitter accounts are fake.

All social media platforms have faced questions regarding the impact bots can make on spreading misinformation since 2016 when Russia interfered in the U.S. election to favor Donald Trump and damage his opponent Hillary Clinton.

Meta (NASDAQ:) owns rival platforms Facebook and Instagram. According to the most recent data, fake accounts account for about 5%. Meta estimates that 11% of fake accounts are duplicate accounts, which is a common practice on Twitter.

Twitter has strict rules against spamming and impersonation. This means that fake accounts can be banned from the platform if they are used to deceive or manipulate people.

CLEARING SPAM

Twitter invested over the years in clearing spam accounts. Smyte was a security and safety company that Twitter purchased in 2018. Twitter deleted “spammy” and “suspicious accounts in an effort to increase its health. In July 2018, its user base dropped by 1 million and the stock fell.

Filippo Menczer, a researcher at Indiana University’s Observatory on Social Media stated that Twitter is more aggressive in taking down fake accounts. However, the threat nature is changing and it is harder to determine.

Menczer stated that “manipulation is also becoming more sophisticated” with well-organized networks and “cyborg accounts”, which can be controlled both by humans and software. He added that bad actors could “flood the system and delete their contents to avoid detection.”

Researchers have found that even though the bots are relatively small in number, they can still have a significant impact on online conversations.

A Carnegie Mellon University study that examined the spread of COVID-19 lies in 2020 revealed that 82% of top tweeters were bots.

Three sources close to the issue told Reuters that Twitter’s measurement and detection spam accounts (or false ones) is a complicated problem. They are not understood or well-understood by most employees.

Twitter has a variety of metrics and definitions for measuring such accounts. This is dependent on Twitter’s ability to detect spam content, according one source. The source also said that it is not able to accurately determine the amount of spam and fake accounts, as new accounts are constantly being created.

Another source said that Twitter’s data and metrics transparency has been “abysmal for ever,” and blamed disorganized management. Plausible denial has proven to be the best route for Twitter’s leadership.