Study of 58 Facebook users' feeds in Georgia found partisan content elbowed out traditional news sites after Facebook reversed its political ad ban
“partisan content quickly elbowed out news sites.” @corintxt's first for The Markup, built on the Citizen Browser project (see credits!) https://themarkup.org/... https://twitter.com/... Adrianne Jeffries / @adrjeffries : More about Citizen Browser here: https://themarkup.org/... Andy Stone / @andymstone : @JuliaAngwin And while we're talking sample size, this bit from the story is especially noteworthy: “20 [Georgia] panelists were not exposed to any election-related content on their feeds as defined by our keyword search.” So this story is actually based on 34 people?! Joshua Lowcock / @jlowcock : The results of @themarkup Citizen Browser project promise to be fascinating https://themarkup.org/... Surya Mattu / @suryamattu : .@corintxt crunched the numbers and found that after Facebook flipped the switch for political ads, partisan content elbowed out reputable news outlets in our panelists' news feeds. https://themarkup.org/... /2 @themarkup : NEW: Ahead of today's crucial Senate runoff elections in Georgia, @Facebook reversed its political ad ban in the state, and the impact was visible on some users' feeds quickly. Read our first #CitizenBrowser story from @corintxt. https://themarkup.org/... See also Mediagazer
The findings immediately drew pushback from Facebook's Andy Stone over sample size, noting 20 of the 58 panelists saw no election-related content under the study's keyword definition. The timing matters: the very next day Facebook said it would reinstate the ban once polls closed, and by March it moved to lift the restriction nationwide.
First-order effects
During the decisive runoff window, the 34 Georgia panelists who did see election-related content got a feed where paid partisan material outranked traditional news sites — exactly the audience Facebook's carve-out was designed to reach.
Facebook's spokesperson challenged the study's effective sample, turning the dispute into a public fight over whether independent audits like Citizen Browser can credibly measure feed composition.
Second-order effects
When Facebook lifts the ban US-wide in March, the Georgia dynamic — partisan spend crowding out news publishers' organic reach during election periods — scales to every contested market at once.
News sites competing for the same attention face a recurring disadvantage each time political advertising reopens, strengthening the case for platforms or regulators to treat feed exposure as something requiring independent verification rather than self-reporting.
Third-order effects
If the pattern holds, accountability shifts to external measurement projects like Citizen Browser, because Facebook's own labeling policy exempts politically funded outlets such as Breitbart — meaning disclosure rules do not correct the imbalance the study documented.
The cycle of imposing, easing, and re-imposing ad bans points toward codified election-period advertising rules, though with the Clarity Act stalled in the Senate, platform policy — not legislation — remains the operative constraint.
The trend: Election-cycle ad bans are becoming episodic dial-turns whose real-world feed effects are documented by independent audit infrastructure rather than the platforms themselves.
Facebook is a newstand. But no one can see which news Facebook is pushing to the top. So we built an app for that called #CitizenBrowser. Our first finding: the sharp impact of Facebook's political ad ban reversal in the Georgia Senate elections. https://themarkup.org/... /1
@JuliaAngwin Interesting that in this whole Twitter thread you don't mention the sample size - 54 people out of a record (for a runoff) 3 million ballots cast. I'm no PhD but I've heard that the plural of anecdote is not data.
We've spent the last ten months building #CitizenBrowser, a project that aims to peek inside the Black Box of social media algorithms, by building a nationwide panel to share data with us. Today, we are publishing our first story from the project. /1
The #CitizenBrowser project seeks to illuminate the content Facebook elevates in its users' feeds. Learn more about how @suryamattu, @leonyin, @angiewaller and @jonkeegan built the standalone desktop application. https://themarkup.org/...
The panel gave us our first peek inside the black box of Facebook's algorithms. In the past month we saw how news feeds in Georgia changed dramatically when Facebook flipped the switch to turn on political ads. https://themarkup.org/... /9
What game is FB playing in Georgia? Why banning political ads in all of the Usa BUT Georgia, where they let disinformation spread wild? https://twitter.com/...
@WillOremus It's remarkable that @facebook hired essentially a Twitter troll to do their comms. Just like the rest of the company, he has zero tact and couldn't care less about appearing as if he's remotely concerned.
Building third-party datasets is essential. FB does everything it can to make this as hard as possible. Some of the measures they took to make this as hard as possible are mentioned in our appendix: https://themarkup.org/... https://twitter.com/...
I developed automated redaction software with @mynameisfiber to scrub personal identifiers (pink) from the FB feed... Turns out FB is all identifiers? https://twitter.com/...
@eric_seufert @mosseri @WillOremus This news story is dog bites man. Prior to a critical election there being a lot of partisan ads is what one would expect. The same occurs with TV ads.
@mosseri @WillOremus Most mainstream criticism of FB is ill-informed and in bad faith. FB is certainly deserving of rebuke for certain things but most of whats written on social media is just frivolous pile-on
There's this analysis going around looking at FB feeds of 54 panelists from Georgia before and after Facebook lifted their ad ban. The point is obviously causal inference about the effect of Facebook's ban being lifted... but there's really no reason to find this credible. https:…
Researchers paid Facebook users in Georgia to monitor everything in their news feeds. https://themarkup.org/... Several fascinating insights, but one of the most surprising is the low engagement D-leaning content earns vs. R-leaning (despite Ds massively outspending Rs). https://…
@andymstone @JuliaAngwin Respectfully, Facebook has this data and does not make it publicly available. We'd all love to see it. Why not share it if you think the conclusions are flawed?
If I was a Facebook PR I would simply try not to draw attention to the fact that I'm being evil and stymieing the ability of reporters to try and throw light on my shady business practices but that's just me!
Whenever reporters and researchers try to understand how Facebook works, employees say the data is incomplete. The Markup explained the limitations in the story, which was a small sample size. https://twitter.com/...
One fascinating tidbit in this is how *little* political content there was: over A THIRD of the (admittedly small) sample saw NO election-related content on their feeds in December! https://themarkup.org/...
Today at @themarkup we published our first story based on #CitizenBrowser data. In it, we found that Facebook ads had a big impact on political news seen by Georgians ahead of the crucial Senate runoff. Here's a quick look at how the story came together. https://themarkup.org/...
In addition to be a cool project by cool people, it's great to see a project really dig in on their methodology. Citizen science projects owe it to the people who contribute their data to explain how they're using it, and @themarkup is doing a heck of a job of that here. https://…
Really amazing initiative from @themarkup - first time we can see, in close to real time, how policy changes show up in user feeds. Congrats @JuliaAngwin and team https://twitter.com/...
sooner or later, we will all be talking about this and/or similar projects inspired by this one. What a GREAT use of human and technological resources. @themarkup is just on a whole 'nother level. https://twitter.com/...
My favorite part of @themarkup's great work on Citizen Browser so far is this brutal and thorough breakdown of all the ways Facebook has made it harder for researchers to examine the platform. https://themarkup.org/...
Some excellent journalism being done here. Deep tech reporting is crucial for understanding how technology can influence our lives, governments and economies (and help inform the ways it needs to be regulated). https://twitter.com/...
When Facebook began reallowing political ads in Georgia, campaigns and partisan PACs reacted quickly, spending millions of dollars on sponsored posts in just two or three weeks. An investigation by @themarkup found partisan content elbowed out news sites. https://www.niemanlab.or…
We didn't want to build Citizen Browser, but we sorta had to if we wanted to get a glimpse of what real Facebook users are seeing in their feeds. Read @corintxt's 1st story using this data: https://themarkup.org/... AND be sure to read our methodology: https://themarkup.org/...
Facebook withholding data from the public and reporters then dunking on reporters who try and get some of that data for small sample sizes is... a brave PR strategy https://twitter.com/...
It's been their model. And it's how reporters have responded to people But more importantly it's how both have responded to marginalized people . Deride them for lack of access you've denied them https://twitter.com/...
Buried in this is $450 million of advertising in Georgia run-off, Facebook being out of control due to scale, Facebook carrying zero liability for its amplification decisions of harmful content, Facebook microtargeting harmful content to vulnerable people. https://twitter.com/...
Despite investments in “sophisticated teams and systems,” over and over again, @facebook failed to enforce their policies and allowed a Super PAC to advertise #misinformation in the lead-up to today's #GeorgiaRunoffs. https://popular.info/...
Facebook flipped the switch on political ads in Georgia, and it completely changed what users saw in their feeds — “partisan content quickly elbowed out news sites.” @corintxt's first for The Markup, built on the Citizen Browser project (see credits!) https://themarkup.org/... ht…
@JuliaAngwin And while we're talking sample size, this bit from the story is especially noteworthy: “20 [Georgia] panelists were not exposed to any election-related content on their feeds as defined by our keyword search.” So this story is actually based on 34 people?!
.@corintxt crunched the numbers and found that after Facebook flipped the switch for political ads, partisan content elbowed out reputable news outlets in our panelists' news feeds. https://themarkup.org/... /2
NEW: Ahead of today's crucial Senate runoff elections in Georgia, @Facebook reversed its political ad ban in the state, and the impact was visible on some users' feeds quickly. Read our first #CitizenBrowser story from @corintxt. https://themarkup.org/...
Facebook banned political ads in the US, EXCEPT in Georgia, and they still allowed 100 ads with election misinfo? This isn't accidental, it's a willful refusal to protect democracy. https://www.washingtonpost.com/ ...