Researchers: Trump tweets with fact-check labels spread further on Twitter than those without; tweets blocked by Twitter remained popular on Facebook and Reddit
We analyze the spread of Donald Trump's tweets … NYU : Despite Warning Labels, Trump's Election Misinformation Tweets Spread Widely Across Social Media Platforms, New Study Finds Justin Hendrix / Tech Policy Press : New research points to role of social media in stoking division in U.S. Andrew Wyrich / The Daily Dot : Trump's false election fraud tweets spread even more with Twitter's warning label, study finds Tweets: @csmap_nyu : Last fall, Twitter attached a warning label or blocked engagement on hundreds of Trump's election misinformation tweets. Our new HKS Misinformation Review paper finds these posts continued to spread widely — on Twitter and other platforms. 🧵 1/ https://misinforeview.hks.harvard.edu/ ... Jason Kint / @jason_kint : Interesting study further supporting benefits of hard inventions that don't remove any posts but instead turn off velocity and reach by preventing a post from being shared into other timelines. https://twitter.com/... Swapneel Mehta / @swapneel_mehta : Interventions are not straightforward tools to deploy; we need to put a lot more thought into the consequences of even well-intentioned policies meant to limit misinformation! https://twitter.com/... Brooke Binkowski / @brooklynmarie : Twitter flagged Donald Trump's tweets with election misinformation: They continued to spread both on and off the platform | HKS Misinformation Review https://misinforeview.hks.harvard.edu/ ... Wendy Via / @wendyvia : SM policy enforcement needs to be more comprehensive. Warnings are not the same as stopping disinformation. https://twitter.com/... @nyuniversity : Messages with warning labels spread further and longer on Twitter than did those without labels, according to a new @CSMaP_NYU study of 1,149 of Trump's tweets. And messages that Twitter blocked entirely went viral on Facebook, Instagram, and Reddit. https://www.nyu.edu/... Michelle M. Shafer / @michelleshafer : Stands to reason. What's the first thing people did when someone tried to “recall” an email message, back when that was a thing? Yep, read it and try to figure out the reason for the recall! Albums back in the 80s marked with “explicit lyrics” warnings also sold really well. https://twitter.com/... @usatodaydc : An NYU study of Trump tweets raises new questions about the ability of social media companies to halt the flood of falsehoods during election cycles. https://rssfeeds.usatoday.com/ ...
Context & Ripple Effects
The NYU CSMaP paper lands as a verdict on the labeling strategy platforms bet on during the 2020 election: last fall Twitter attached warning labels to hundreds of Trump's election tweets and blocked engagement on others, a move contemporaneous coverage called a public-relations win that was "too little, too late" as a product decision (warning labels during the election). The study now measures what those interventions actually did to spread.
It also fits an arc of measurement-first research from the same group — an earlier NYU analysis found misinformation-heavy publishers drew six times more engagement than trustworthy sources like WHO over the same August 2020–January 2021 window — and it quantifies the feedback loops between Trump and his influencer-and-follower base that reporting flagged going into the election.
First-order effects
- On Twitter itself, the intervention backfired at the margin: tweets carrying fact-check labels traveled further and longer than unlabeled ones, meaning the label functioned as no deterrent for the audience doing the sharing.
- Tweets Twitter blocked outright did not disappear — they stayed viral on Facebook, Instagram, and Reddit, so enforcement on one platform simply shifted the same content's audience onto platforms that hadn't acted.
Second-order effects
- Facebook's lighter-touch approach — labeling politicians' voting posts with generic "Get official voting info" notices rather than fact-checks (its July 2020 policy) — left it structurally positioned to absorb traffic that Twitter's harder line pushed out, rewarding the least restrictive policy in the set.
- Platform teams designing moderation now have counter-evidence that visible friction can amplify reach within the enforcing platform, complicating the label-as-default playbook both companies had adopted.
Third-order effects
- If blocking on one network merely reroutes virality to adjacent ones, effective containment requires cross-platform coordination no single company controls — pushing the debate toward distribution-layer liability rules rather than voluntary per-platform labels.
- The longer pattern holds past the bans themselves: since January 6, Trump and allies have beaten down filtering efforts even after his removal from these platforms, so the study documents why removal-without-coordination underperforms while the political will to re-filter erodes (the post-ban rollback of election-lie filters).
The trend: Single-platform content moderation is being exposed as leaky containment, with measured-spread research driving the argument toward coordinated or regulator-imposed distribution-layer standards.