Data from Facebook users: 12 political groups were in the top 100 recommended in Jan.; FB said it would not recommend “political content or social issue groups”
According to Citizen Browser data, the platform especially peppered Trump voters with political grouprecommendations
The Markup
Context & Ripple Effects
The arc here is a pledge colliding with measured reality. Facebook confirmed in October that it had stopped recommending political and social-issue groups ahead of Election Day [[a:1160789]], and Zuckerberg followed in late January by making the ban permanent [[a:962555]]. The company's own internal research had already flagged the problem: roughly 70% of its most active civic Groups were judged too toxic to recommend [[a:962658]].
What makes this Markup story matter is who caught the gap. Citizen Browser — the same panel project behind the Split Screen comparison of Trump and Biden voters' News Feeds — found 12 political groups still sitting in the top 100 recommendations in January, disproportionately aimed at Trump voters. An independent panel is now auditing whether Facebook's civic-content promises survive contact with the actual recommender.
First-order effects
Trump voters on the platform were still being served political group recommendations in January, directly contradicting the stated policy Facebook had announced in October and reaffirmed through Zuckerberg's permanent-ban pledge.
Second-order effects
Citizen Browser's panel becomes the de facto verification layer for Facebook's civic-content commitments — every future pledge about News Feed or group curation is now testable against observed feeds, raising the cost of policy-enforcement gaps.
Third-order effects
If audit-style projects keep documenting divergence between announced curation policy and delivered recommendations, platform accountability shifts from self-reported statements toward independent measurement — a structural move with regulatory implications for how algorithmic distribution claims are policed.
The trend: Platform curation promises are shifting from trust-based assurances to independently audited delivery, as citizen-panel data projects begin measuring what recommenders actually serve.
After the riot in the capitol, Facebook pledged it was going to stop recommending users join political groups. Here's what the platform's algorithm did instead https://themarkup.org/... https://twitter.com/...
@alfredwkng
@alfredwkng on x
New from me and @LeonYin: 6 days before Election Day, Zuckerberg told the Senate that Facebook stopped recommending political groups. They reiterated this after the Capitol riots. We found recommended groups with conspiracy theories and death threats: https://themarkup.org/...
@fightfortheftr
@fightfortheftr on x
We need an immediate moratorium on Big Tech companies practice of non-transparent algorithmic amplification of posts and groups https://twitter.com/...
@yathinkn
@yathinkn on x
Oh sounds like the Australian Government and Facebook have a lot in common. Make an announcement and change something superficial. Except rarely do journalists - with large platforms - reveal that nothing came from the announcement 😞 https://twitter.com/...
@freepress
Free Press on x
Facebook has a history of announcing improvements to its platform that later fall flat. “Their policy looks good on its face but they're narrowly interpreting that policy so it doesn't apply when things really get hot,” says @JGo4Justice. https://themarkup.org/...
@juliacarriew
Julia Carrie Wong on x
Facebooks's Group recommendations have long been one of its most dangerous and opaque products. So it's very exciting that @themarkup is now auditing them, though the results are unsurprisingly disturbing. https://themarkup.org/...
@jayrosen_nyu
Jay Rosen on x
From to time I share newsroom projects under the heading: “Three of the most powerful words in journalism: help us investigate.” The @themarkup's “citizen browser” project may be the best example yet. (They pay people to participate.) https://themarkup.org/...
@capaction
CAP Action on x
Despite Facebook's claims to the contrary, “the platform continued to recommend political groups to its users throughout December...[and] also continued to recommend political groups throughout Jan, including after it renewed its promise not to on Jan. 11.” https://themarkup.org/…
@kashhill
Kashmir Hill on x
“Facebook Said It Would Stop Pushing Users to Join Partisan Political Groups. It Didn't.” (Because it only stopped recommending groups that checked a box self-reporting themselves as political.) https://themarkup.org/...
@hankgreen
Hank Green on x
In fairness to Facebook, if they had stopped playing this active role in destabilizing our society and radicalizing people, they would have made a little less money. https://themarkup.org/... https://twitter.com/...
@shiraovide
Shira Ovide on x
What percentage of stories about Facebook are basically: Facebook said it doesn't allow X or stopped doing Y. OK, but actually...here is X on Facebook in spite of its policies, and here is Facebook doing Y. https://twitter.com/...
@adrjeffries
Adrianne Jeffries on x
Trump voters were targeted hardest by Facebook political group recommendations https://themarkup.org/...
@alfredwkng
@alfredwkng on x
These are groups that Facebook recommended throughout December that called to storm the Capitol, with members who were in DC on Jan. 6, and claimed Ashli Babbitt was still alive. Facebook deleted this post after we reached out to them. https://themarkup.org/... https://twitter.co…
@alfredwkng
@alfredwkng on x
Also, just look at how likely you were to be recommended a Facebook group depending on how you voted. We found a significant discrepancy between Trump and Biden voters: https://themarkup.org/... https://twitter.com/...
@leonyin
Leon Yin on x
Among the top 100 Groups FB's rec sys pushed to the 508 Trump voters in our panel: calls to violence against elected officials, conspiracy theories, and logistics for attending DC on the day of the insurrection. https://themarkup.org/...
@aaschapiro
Avi Asher-Schapiro on x
In October, Facebook promised its recommendation algorithm would stop pushing users to join “political groups.” In the ensuing months, FB continued to recommend people join a “Candace Owens for POTUS,” group which encouraged members to “StormTheCapitol” https://themarkup.org/...
@colinlecher
Colin Lecher on x
Facebook said it had stopped recommending political groups to users. But we found they recommended them to up to 20 percent of Trump voters: https://themarkup.org/...
@leonyin
Leon Yin on x
We analyzed Facebook Group recs (fb[.]com/groups/discover) for +1.9K users throughout Dec. We found 12 political groups in the top 100 most commonly rec'd Groups. We ran the numbers again by how users voted in 2020... https://twitter.com/...
@juliaangwin
Julia Angwin on x
In the fall, Facebook promised to stop recommending political groups. It renewed its promise again after the Capitol riots. But guess what? They didn't do it, according to data collected from our #CitizenBrowser project. @LeonYin & @alfredwkng report: https://themarkup.org/...
@evan_greer
Evan Greer on x
What's frustrating is that this is actually one of the promises Facebook could keep and would actually make a difference, unlike when they promise to remove all content about a specific topic or espousing a specific ideology, which is essentially impossible at scale https://twitt…
@sivavaid
Dr. Siva Vaidhyanathan on x
The pattern continues: Facebook promises a policy change. It does something superficial. Users adjust. Journalists reveal that Facebook failed to change it at all. https://twitter.com/...
@danieljkelley
Daniel Kelley on x
FB continued to recommend political groups after saying it wouldn't, according to @themarkup's citizen browser investigation. As always, god bless @themarkup- so crucial to have robust transparent investigations of digital social platforms. https://themarkup.org/...
@themarkup
@themarkup on x
Zuckerberg promised Facebook would stop recommending “political content or social issue groups” to prevent the spread of misinformation and violence coordination. That promise was broken, according to #CitizenBrowser data. @alfredwkng and @leonyin report. https://themarkup.org/..…
@farai
Farai Chideya on x
From @themarkup: Facebook Said It Would Stop Pushing Users to Join Partisan Political Groups. It Didn't. According to Citizen Browser data, the platform especially peppered Trump voters with political group recommendations https://themarkup.org/...