Oct. 25 (UPI) -- Twitter's crackdown on automated "bots," spam and fake accounts resulted in the loss of 9 million users in the third quarter, the company said Thursday.
The social media giant said it anticipated the loss after it started using better software to pinpoint questionable accounts or prevent them from signing up at all.
"We made progress preventing spammy or suspicious new account creation by requiring new accounts to confirm either an email address or phone number when they sign up to Twitter, and we improved the detection and removal of previously banned accounts who attempt to evade suspension by creating new accounts," Twitter said in its quarterly filing.
With the losses, Twitter's user base fell to 326 million -- which, after two straight quarters of losses, roughly matches the level it was at in early 2017. Twitter said it added new users, but wiped out more than it gained. Daily active users grew 9 percent over the last year, but they account for fewer than half of monthly users.
"We do see health [of the platform] as a growth vector over the long term," Twitter CEO Jack Dorsey said on an earnings call Thursday. "This is an extremely important initiative for us, not only for the experience of Twitter, but we believe the long-term growth of the platform."
Twitter said it expects to lose several million more users in the fourth quarter. It also reported a profit for the last four quarters, something it hasn't achieved before.
Revenue growth was up 29 percent year-over-year and (GAAP) net income totaled $106 million. Video accounts for half of the company's ad revenue.
Twitter also addressed how it's preparing for the Nov. 6 midterm elections so users see the best, most informed content first.
"Protecting the integrity of elections is another essential part of improving the health of the public conversation on Twitter," Twitter said. "After launching the Ads Transparency Center in Q2, we made a number of additional updates in Q3 in preparation for upcoming elections, including provider clearer guidance around attributed activity."