Twitter is inserting tweets into the feeds of users that do not follow the accounts which posted them, sometimes inadvertently amplifying extremist rhetoric
Oliver Darcy / CNN : Tweets: @oliverdarcy , @clarajeffery , @oliverdarcy , @dduart3 , @jenmercieca , @alexhern , @chuckrossdc , and @cwarzel See also Mediagazer Tweets: Oliver Darcy / @oliverdarcy : Some of the people I've noticed Twitter has been amplifying: Diamond & Silk, Bill Mitchell, Charlie Kirk, James Woods, Candace Owens http://www.cnn.com/... http://twitter.com/... Clara Jeffery / @clarajeffery : they worked the refs, yet again. or Dorsey and his team are sympathetic to white nationalists. Take your pick. http://twitter.com/... Oliver Darcy / @oliverdarcy : Of course, the irony of Twitter inserting some of these right-wing voices into feeds of users who do not follow them is that these are the people who keep accusing Twitter of “shadow banning” them. It's really...the opposite? http://www.cnn.com/... http://twitter.com/... Daniel Duarte / @dduart3 : 2017: Social media encourage filter bubbles and echo chambers! They need more diversity! *algorithms do that* 2019: Social media push people to extremism! They need to police radical speech! http://twitter.com/... Dr. Jennifer Mercieca / @jenmercieca : Fascinating case study. The algorithms are designed to push engagement, to get us to post our most extreme takes in order to be promoted. Neutral, centrist, reasonable positions? Demoted. This is the real untold scandal. Express yourself with sensationalism and you'll benefit. http://twitter.com/... Alex Hern / @alexhern : Twitter, jealous of all the attention Facebook is getting, tries its own hand at algorithmic radicalisation http://edition.cnn.com/... Chuck Ross / @chuckrossdc : Stop clicking on their tweets and this probably won't happen anymore http://twitter.com/... Charlie Warzel / @cwarzel : i think what this speaks to is how pretty much any tool that helps promote or boost user engagement is going to end up having the unintended consequence of amplifying misinformation/conspiracies/ divisive rhetoric. http://twitter.com/... See also Mediagazer
Context & Ripple Effects
Twitter's decision to surface tweets from accounts users never opted into follows a playbook already documented elsewhere: at YouTube, former extremists describe being radicalized by an algorithm that kept surfacing extremist content, and later research suggests cross-viewpoint exposure in feeds can deepen rather than reduce political polarization.
The CNN report lands amid growing scrutiny of how feeds are curated — Facebook's own News Feed team bristled when Kevin Roose's CrowdTangle engagement rankings exposed what its algorithm rewarded, and former employees later described deliberate tests of changes tilting toward right-leaning outlets. What makes this round different is that Twitter's amplification appears inadvertent, surfacing figures like Diamond & Silk, Bill Mitchell, and Candace Owens to audiences that never followed them.
First-order effects
- Accounts named by Oliver Darcy — Diamond & Silk, Bill Mitchell, Charlie Kirk, James Woods, Candace Owens — gain reach beyond their follower base without paying for it, while users who never opted in are exposed to rhetoric they did not choose.
- Twitter's own executives face immediate credibility pressure, since the company's stated position that its timeline reflects user choices is contradicted by its own product behavior.
Second-order effects
- Advertisers and brand-safety partners gain fresh evidence that algorithmic placement, not just organic following, puts brands next to extremist content — the same exposure problem that made Facebook's feed team defensive about engagement data.
- Competitors curating feeds algorithmically, notably Facebook, inherit the scrutiny: any claim that ranking is neutral gets tested against the Twitter example.
Third-order effects
- If recommendation systems keep distributing content past the follower graph, creators' incentive shifts from building audiences to engineering algorithmic pickup — the escalation dynamic experts describe among trolls who built careers on Twitter and now push shock tactics harder as normal-user audiences shrink.
- Algorithmic amplification becomes a regulatory target distinct from content moderation: the question moves from 'what did the platform host' to 'what did the platform distribute,' forcing platforms to treat ranking choices as accountable editorial acts.
The trend: Engagement-optimized feeds are dissolving the follower graph as the boundary of distribution, turning algorithmic amplification itself into the platform's most consequential — and least governed — editorial decision.