Sebastian JamesKey takeaways
- A few kinds of account carry most of the rumor traffic. The 817 accounts that mostly share links (17% of 4,870) put out 68% of the posts in the rumor stories tracked.
- Link sharing is the behavior that sets them apart. The groups that mostly share links are 16% to 29% likely to have carried a rumor story. The groups that write their own posts are 1% to 11%.
- Most of the crowd never touches a rumor. 61% of accounts are casual posters who write a couple of posts, and 1% of them ever carried one.
- These are behavior types, not motives. We can say how an account posts. We cannot say who runs it or why, or whether it is a person.
| 68% | 4,870 | 6 |
|---|---|---|
| of the rumor-story posts come from the 17% of accounts that mostly share links | accounts measured, from 26,932 posts in the FC26 archive | composite personas, no handles |
The data comes from FC26, Michigan United's election early-warning program, which archives public election posts through Open Measures. Counts only. No handle appears anywhere in this piece.
How we built them
A persona here is a composite. We took every account in the archive with two or more posts, 4,870 of them, and measured how each one posts across everything it has posted: how much, over how many days, how often it shares a link, how often it repeats text other accounts also post, how often it replies, how often it is among the first to post a story, and the hours it posts. We sorted the accounts into groups by volume (a couple of posts, a few days of posts, a daily poster) and then by posting style within each. Rumor involvement was measured afterward, from the stories our claim gate kept as checkable claims about how an election is run, and was not used to choose the accounts.
Handles are scrambled inside the database before any of this runs, and each persona is an average of many accounts, so it describes no one.
The six personas
Occasional Relay: 524 accounts
It drops a link or two and moves on: 2 posts in the archive, 91% of them links, usually with little or no comment. It is found mostly on the Fediverse, Gab, Disqus and Truth Social, and it is often among the first to post a story. It put out 33% of the rumor-story posts, and 16% of its accounts carried one.
Regular Link Sharer: 241 accounts
It shares links on a steady rhythm: 7 posts over 5 days, 89% of them links, and it asks a lot of questions. It is on the Fediverse and Gab more than anywhere else. It put out 25% of the rumor-story posts, and 17% of its accounts carried one.
Heavy Link Sharer: 52 accounts
It shares links on a steady rhythm: 34 posts over 17 days, 96% of them links, and it asks a lot of questions. This group is smaller but runs like a standing channel. It put out 9% of the rumor-story posts, and 29% of its accounts carried one.
Regular Plain Poster: 912 accounts
It writes its own posts in a handful of sittings, 7 over 5 days, mostly on Bluesky. 2% of them are links. Its rumor posts include more claims that voting is unsafe than the link groups’ do, though that rests on only 10 posts. It put out 12% of the rumor-story posts, and 4% of its accounts carried one.
Heavy Plain Poster: 150 accounts
It writes its own posts every few days, 29 of them over 15 days, nearly all on Bluesky and Disqus and mostly about named people. It rarely shares links. It put out 11% of the rumor-story posts, and 11% of its accounts carried one.
Occasional Plain Poster: 2,991 accounts
It is most of the crowd: 2 posts over 2 days, almost never a link, mostly on Bluesky. Only 1% of its accounts ever posted in a rumor story. It put out 9% of the rumor-story posts.
Who carries the rumors
The gap between the grey and red bars is the story. The Occasional Plain Poster is 61% of all accounts and only 9% of the rumor posts. The Occasional Relay is 11% of accounts and 33% of the rumor posts.
Where they post
The link groups live mostly on the Fediverse, Gab, Disqus and Truth Social. The plain posters live mostly on Bluesky and Disqus. Reddit is not in this archive, because Open Measures does not carry it, so these personas describe the platforms we can see.
What this suggests for answering rumors
- Answer the link, not the account. The Occasional Relay posts once or twice and is gone. Answering each one cannot keep up. A clear answer on the page the link goes to, or at the claim’s source, reaches everyone who follows it.
- Expect a steady channel from the heavy link sharers. There are few of them and they post for weeks. A standing, regularly updated answer fits that better than a one-off reply.
- The plain posters are mostly ordinary people. Their rumor posts include some claims that voting is unsafe, a small base. For them, a trusted local voice and a plain “here is what is true in your county” works better than a correction aimed at a link.
These are suggestions from the pattern, not tested results.
What a persona cannot tell you
- Who is behind an account, or why. A link relay can be a person, a bot or an organization. We measure behavior only.
- Whether accounts are coordinated. We count accounts that repeat each other’s text, but repetition is not proof of coordination.
- More than the archive shows. The rumor-story posts are 332 posts in 34 stories, a thin base. The persona list also shifts a little as the data grows: a small group of 20 accounts that copied text heavily on Oct. 22 has since merged into the larger Regular Link Sharer group. The split between link sharers and plain posters held in both runs.
Limits. Archive: Bluesky, Disqus, Fediverse, Gab, Gettr, Truth Social, Telegram and 4chan through Open Measures; no X, Facebook, Reddit, YouTube or Instagram. Personas are rebuilt from the data and named by us (volume tier plus posting style); the names are working labels. Reply counts are thin because many posts do not record what they replied to. A person approves every public version.
What shape is the conversation about your race?
Digital Strategy Group runs election narrative scans for candidates, campaigns and organizations. We show you what is moving, how far and what to do about it. Or call or text 773-879-3538.
Email me about a narrative scan