Ukraine can't bomb all the bad posts away

Our next live show at Baby’s All Right in Brooklyn on October 27th. It’s a night of comedy, music, some journalism, and, of course, a costume contest. Whoever shows up in the most brainrotted outfit wins a big prize. We also have not one, but two very special guests: Annie Rauwerda, the genius behind Depths of Wikipedia, and comedian Alex Pearlman, better known as pearlmania500. Grab tickets while you still can!!!

This week, Ukrainian forces bombed three data centers used by Yandex, Russia’s largest tech company and most popular search engine. The bombings have caused outages across all of Yandex’s services, from cloud storage to its rideshare app. But it’s also had the effect of silencing inflammatory bot accounts on social media. Or has it?

Ever since the bombings, the story that Ukraine knocked out Russian bots has spread everywhere, from news articles to politicians to Ukrainian President Volodymyr Zelenskyy himself. But a few days later, the official sources started backtracking on the story, as researchers noticed that none of the initial evidence — mostly from random posts on social media — held up to scrutiny. Which begs the question, how did the story spread so far, so fast, if it was completely fake?

Well, we’ve tried to get to the bottom of this and we can say that there are some signs that bot activity may be down. Though it may not be because of the drone strikes.

The earliest example we can find of a claim that the strikes on Yandex had reduced bot activity was on October 8th, the same day as the first strike. Jeremiah Fowler, an independent cybersecurity researcher, posted on X that Cloudflare had reported a drop in bot traffic from Yandex since the strike. As the bombings continued, these claims would get viral reposts and be reported as news. It culminated on October 12th, when Zelenskyy delivered a speech saying, “It’s pretty clear where those bots were coming from, if they disappeared when Yandex went offline.”

The problem is, it’s not so clear once you look at the data. Zelenskyy later qualified his statement, since as other sources have reported, what Cloudflare calls “Yandex bot traffic” refers to the web crawlers for its search engine and other services. Automated accounts on social media are completely different from what Cloudflare was tracking, and they’re a lot harder to track in the same kind of simple chart. There are a lot of reasons for this, but the biggest is that it’s gotten harder to study social media activity at all, anywhere. Which is why we have to pay attention to some more anecdotal information.

There were wider reports, mostly on X, that people were experiencing decreased bot activity. Many accounts and news reports debunking the Cloudflare story had to note that there could be a separate reason people were seeing fewer bots online, it just wasn’t related to the strikes on Yandex. And we think we’ve found that other reason.

X recently announced that it was implementing liveness checks on its accounts. A liveness check is some kind of proof that an account is being run by a physical human with a body — usually a short webcam video or a biometric pass. They’re becoming increasingly popular in the age of autonomous AI, which can now consistently beat other tests like CAPTCHAs (a moment of silence for Lord Inglip).

We have evidence that X has been using liveness checks, but this is the first public announcement, suggesting it’s now being used on many more accounts — and possibly catching a lot more automated accounts. When were the checks rolled out? According to the page source on X’s Help Center, it was on October 8th, the same day as the first Yandex data center was bombed.

Is that really what happened here? One enormous coincidence, where a bot purge just so happened to line up with a news story that speaks to people’s oversimplified understanding of bots? We can’t be certain, since we have no way of knowing how effective the liveness checks have been. 

X has severely restricted most of its public analytics data, and its press team hasn’t moved very far from their “💩 autoreply” days. They didn’t respond to a request for comment on this story, but X did say on Thursday that it deleted over 13,000 accounts that were associated with “foreign influence operations” meant to change political conversation around the US midterm elections. Which is likely the larger reason why people are noticing a change, even outside of the Everything App. 

Every major social network has spent years tweaking and refining its policies around political content during election season, and a bunch of major elections are being held around the world. Yesterday, Brazil’s Supreme Court ordered Meta to change its algorithms in advance of next week’s presidential election runoff. Thanks to algorithms and automatic translations globalizing our feeds, users everywhere will feel the effect of the many changes to those feeds that happen every election season. Especially if those users have been swimming in the sewer of online politics for long enough they’ve forgotten what fresh air is like.

The idea that all the bad political opinions people see online are Russian bots has persisted for more than a decade now. But bombings outside Moscow can’t blow away all the posts you don’t like. And that’s doubly true on global platforms that are now openly allied with authoritarian governments. The call is coming from inside the house. We’ve known for years who’s manipulating what you see online, and they’re not running server farms in Siberia, they’re running the apps.