Grilled Cheese

ExploreLog inSign up

Kate Starbird

@katestarbird.bsky.social

3k Following93.8k Followers

Researcher of online rumors & disinformation. Former basketball player. Prof at University of Washington, HCDE. Co-founder of the UW Center for an Informed Public. Personal account: Views may not reflect those of my employer.

PostsRepliesMedia
@noupside.bsky.socialOct 14, 2026, 8:25 PMReposted by @katestarbird.bsky.social

Did Ukrainian drone strikes really silence Russia's social media bots? (No) A real Cloudflare chart fueled a viral story ab MAGA and AfD accounts going quiet. But the evidence doesn't support the rumor. I wrote about bots, vibes, & what we've lost as platforms have shut down tools & data access.

Photo, no description
@texastribune.orgOct 14, 2026, 1:47 PMReposted by @katestarbird.bsky.social

Voting rights groups are suing Texas after the state failed to process more than 500,000 voter registration applications submitted through the Texas Department of Public Safety’s web portal.

@benrobins.bsky.socialOct 14, 2026, 3:33 PMReposted by @katestarbird.bsky.social

It's accepted that there are troll/bot accounts creating fake engagement on X, Facebook etc. but the accounts themselves reside on the servers of Meta, X, etc. Orchestration of fake accounts can therefore be done anywhere connected to the internet, and it's not difficult to relocate such a workload

@katestarbird.bsky.socialOct 14, 2026, 3:36 PM

Evergreen? Our team has been studying election rumors for a few cycles now. We have a "threat framework" for quickly assessing whether an emerging rumor is likely to go viral. We initially published this in 2022, but we're still using it to guide our work: www.niemanlab.org/2022/10/elec...

@katestarbird.bsky.socialOct 13, 2026, 10:45 PM

Sure. I’m not saying that there’s no signal there, but there’s no strong signal. Additionally, there’s now media coverage & commentary explaining how social bots just don’t work that way (X accounts don’t live on a server in Russia) and the original amplifier of this rumor has retracted & apologized

@katestarbird.bsky.socialOct 13, 2026, 10:13 PM

One of the more interesting elements of this work is how Grok is repeatedly called into these live election rumors, & sometimes does a decent job. Go to our interactive graph, select "Size by Follower Count" and you can see/read/link to all of the Grok replies: faculty.washington.edu/kstarbi/Spot...

Photo, no description
@katestarbird.bsky.socialOct 13, 2026, 10:09 PM

If you'd rather follow us directly from our UW site, you can see our work here: www.cip.uw.edu/2026/10/13/e...

@katestarbird.bsky.socialOct 13, 2026, 10:08 PM

Our research team posted a rundown of a few "rigged election" rumors that spread online late last week. We include explanations for how those rumors mislead and an analysis of how they spread on X. Follow/subscribe to our blog for more updates on election rumors: uwcip.substack.com/p/election-r...

uwcip.substack.comElection Rumor Rundown October 13DHS reports misconstrue vulnerabilities as having been exploited, “Ballot Dumping” in Valley Center misleadingly suggest mass voter fraud, and prediction markets already advertising election betting
2
@katestarbird.bsky.socialOct 13, 2026, 8:39 PM

Yes, that's why we chose that term. It's a rough analysis, but it aligns with a bunch of other analyses, new info about the initial graph, and an apology from the guy who popularized the original claim. All converging on the understanding that any effect on social media info ops was super small.

@katestarbird.bsky.socialOct 13, 2026, 8:24 PM

Quite separate from my rough analysis, one of the initial amplifiers of this claim has retracted it and apologized: bsky.app/profile/aida...

@aidanobrien.bsky.socialOct 13, 2026, 6:20 PMReposted by @katestarbird.bsky.social

Notably, Jarosław Wolski, the Polish political scientist who was a primary source of the AfD social media bot claim, has now retracted his statement and apologised (more of this, please!). German journalists have also confirmed the graph was fake and AI-generated.

Apologies and Correction.
I would like to apologize to all Readers for providing false information and a graphic suggesting that bot activity around AfD had dropped by 56%.
According to the findings of @niebezpiecznik, which cites the opinion of the Cemas research institute, the Hasso Plattner Institute, and the
"Polisphere" think tank, there has been no permanent and measurable decrease in pro-AfD posts. Moreover, the original graphic in German turned out to be Al-generated.
Unfortunately, my post with the disinformation reached millions of views, for which I would like to apologize to all readers once again. Of course, I take full responsibility for what I did, and simply and humanly, I am ashamed that I misinformed you. I also apologize to the contested opponents who were the first to point out that I was wrong.
To avoid further disinformation, I have deleted the post. From now on, in topics not strictly related to the military, I will adopt the principle of providing citations or original sources whenever possible. Or clearly stating their absence/unclear source of information so that it is visibly evident in the post.
JW.
Rate this translation:.! A German journalist has tracked down the original graphic that sparked a wave of controversial (to put it mildly, untrue) reports yesterday and today about a drop in activity from Russian bots/trolls.
He created it-"24-year-old steward Frank"—and marked it as ...generated by Al (which we already suspected yesterday).
Unfortunately, the politicians who shared this material further failed to keep that marking (intentionally or by accident 1ºr) and didn't indicate the source either.
What's more, the German research institute Cemas, which deals with disinformation, the Hasso Plattner Institute, and the Polisphere think tank have officially X denied observing any drop in the number of posts that could be explained by an attack on Yandex.
"Neither the activity of existing bots nor the rate of creating new ones has declined, as shown by the data we're collecting."
Well, those are data from Germany, though. Meanwhile, in Poland, Minister Siemoniak said that "ABW has noted a radical drop in the activity of Russian bots online. This is a phenomenon observed worldwide after the bombing of three data centers."
But ABW hasn't publicized anything on this matter so far. Here's hoping they do. That would be something!! For now, we know that "worldwide" apparently means everywhere except Germany...
@katestarbird.bsky.socialOct 13, 2026, 6:19 PM

They’re in Europe, sleeping. (It’s US Pacific time, where I grabbed my data.)

@katestarbird.bsky.socialOct 13, 2026, 6:14 PM

Is there a small impact? That’s totally possible. But the stories of a massive silencing of Russian bots making a qualitative difference on social media discourse… are very likely to have been (significantly) overstated.

@katestarbird.bsky.socialOct 13, 2026, 6:03 PM

www.motherjones.com/politics/202...

@katestarbird.bsky.socialOct 13, 2026, 6:00 PM

Could be. Maybe for non-centralized platforms, there could be some hosting of accounts in Russian data centers. For Meta and X and other centralized platforms, it doesn’t make sense (to me) how social bots would be affected.

@katestarbird.bsky.socialOct 13, 2026, 5:53 PM

I use a commercial product called Meltwater.

P@prm.bsky.socialOct 13, 2026, 5:51 PMReposted by @katestarbird.bsky.social

Traffic volume from yandex web crawler machines went down (a webcrawler is usually lower priority workload as it is mostly refreshing pages ). Web crawler are referred to as bots. Nothing to do with social media posts.

@katestarbird.bsky.socialOct 13, 2026, 5:50 PM

Is there a temporary dip on the morning of the 11th? Maybe? But there’s no sharp drop and things return to normal quickly.

@katestarbird.bsky.socialOct 13, 2026, 5:46 PM

The overall signal is low on Oct 10 between 6-8pm but the multi-day pattern suggests that isn’t unusual, and the conversation rises again the next day.

@katestarbird.bsky.socialOct 13, 2026, 5:43 PM

And here’s X posts by minute for posts containing “AfD” on October 10. No sign of a rapid drop in X posts that day or anything weird between 6-8pm PDT. Just a normal daily pattern.

Photo, no description
Older posts
Terms of UsePrivacy PolicyCommunity StandardsHelpGet the app

Grilled Cheese is a product of Village Compute

Version devBuilt at: 2026-10-15 05:56:38 EDT