The Industry

Twitter Made a Small Change to How It Shows Retweets. That’s Bad News for Bots.

SAN FRANCISCO, CA - APRIL 26:  A sign is posted on the exterior of Twitter headquarters on April 26, 2017 in San Francisco, California. Twitter reported  better-than-expected first quarter earnings with revenue of $548 million, compared to analyst estimates of roughly $512 million. Monthly active users to jumped to 328 million, 7 million more than expected.  (Photo by Justin Sullivan/Getty Images)
Twitter is mixing things up a bit. Justin Sullivan/Getty Images

Without any fanfare, Twitter rolled out a new feature globally on Tuesday that changes the way tweets look when embedded on other websites, like in news stories. Now, instead of showing how many times a tweet has been retweeted, the company is displaying a new metric that includes a combination of replies and retweets that shows how many “people are talking about this.” It’s a change the company confirmed to Slate on Tuesday evening, but has not yet announced publicly. It looks like this:

The company started experimenting with new ways to show how many people engaged with a tweet late last year, according to Twitter spokesperson Dan Jackson. “We found that people viewing Tweets off-platform were more likely to engage with them when we focused on providing conversational context,” Jackson said. In other words, if a tweet shows up on a news story read by someone who isn’t a Twitter user, seeing that it’s been retweeted thousands of times might not make much sense. But Twitter believes that knowing how many people are sharing and commenting—that is, “talking about” the tweet—is a clearer indicator of popularity for those who aren’t familiar with how Twitter works.

By being less specific about the how much tweets embedded around the internet are shared, Twitter is in a way downgrading the importance of retweets as an overall indicator of a popularity. It’s also a move that could help counteract the social network’s massive bot infestation; bots are automated accounts that use software to act on their own to tweet, follow, and retweet others. (There are also bots that automatically reply to tweets, but they are less common.)

Russian-linked bots retweeted Donald Trump’s tweet almost 500,000 times in the final weeks before the 2016 presidential election, according to testimony the company provided for the Senate Judiciary Committee last month. The Russian-bot retweets accounted for 4.25 percent of all retweets from Trump’s account between September 1 and November 15, 2016. To put that in perspective, Twitter said Hillary Clinton received nearly 48,000 retweets from Russia-linked bots during the same time period, meaning the accounts retweeted Trump 10 times more than Clinton in the run-up to and days immediately after Election Day. Last month, Twitter also said its removed more than 50,000 Russia-linked bots as part of its larger effort—along with Facebook and Google—to purge Kremlin-backed accounts that flooded U.S. social media networks around the 2016 election that were being used to spread divisive messages and confuse voters on some of America’s most polarizing social and political issues, like gun control, racial profiling, and Donald Trump.

When thousands of Twitter bots retweet a missive, particularly a political one, it can give the impression that the message has larger grassroots support than it actually does, since a retweet generally indicates that someone found the content of the post interesting or important enough to spread to their followers. If a large number of the retweets aren’t coming from a real person and is then being embedded on a news website, the result is misleading. When a tweet appears to have been retweeted thousands of times, it’s reasonable to conclude the sentiment is shared by thousands of others.

For its part, Twitter has said that it’s making strides in cleaning up its bot problem. It now blocks 523,000 suspicious logins every day for being automated, the company said in a January blog post. Twitter says it’s become better at spotting things like when a bot replies faster to a tweet than a human could and when multiple tweets are suspiciously timed. Some extra thoughtfulness in tackling this problem is important, because not all automation on Twitter is bad thing. Artists make bots that tweet from archives and pull together strings of words to make poetry, not the mention bots that live tweet events, such as information about earthquakes as they’re reported by the United States Geological Survey for places that are at risk, like the Bay Area and Los Angeles.

With this latest move, beating back bots probably isn’t the main motivation. It’s likely the idea is simply to use less language that only people who already use Twitter would understand. But even so, removing the retweet count on embedded tweets that end up in news stories and replacing it with a broader metric does take away some of the bots’ raison d’etre—at least until they figure out how to game this measurement, too.