Editor’s Note: MRC got the idea for its research utilizing xAI Grok from a query posted by an X account @The1Parzival. The published study omitted a hat tip.
Elon Musk has made strides to turn X (formerly Twitter) into a free speech platform but there are still remnants of the Old Twitter censorship regime and they are interfering in the 2024 election. It now appears that X is actively boosting content posted by Democrat congressional members’ accounts, while simultaneously de-amplifying their GOP counterparts in the middle of an election year.
X’s algorithm is actively boosting Congressional Democrats’ content in a chilling display of election interference. Using X’s AI chatbot, Grok, MRC was able to peer into X’s algorithm to see how it was treating all members of Congress. What MRC found was that X gave a distinct advantage to Democrats in both the House and the Senate over their Republican counterparts.- The blatant bias in X’s algorithm reveals an internal revolt against Musk within the company’s ranks. With the 2024 election just months away, MRC’s evidence suggests that a radical remnant is still working against Musk in secret by tilting the scales in favor of the left. MRC had uncovered a spike in censorship cases following Musk’s acquisition of Twitter in 2022, which was the first red flag that holdovers from Old Twitter were actively undermining him and his free speech vision for the platform. The censorship practices of Old Twitter that were thought to be a thing of the past are very much active today.
- The more right-leaning a congressional member, the worse their X visibility score is. MRC discovered a glaring continuum within X’s algorithm. Generally speaking, the more left-leaning a congressional member is, the more visible their account is to X platform users. On the flip side, the more right-leaning a congressional member is, the more their account is penalized by the algorithm.
- Leftist Rep. Maxine Waters (D-CA) is considered the highest-ranked House member by X’s algorithm. The way the X algorithm applies its criteria to determine which accounts to boost is nothing short of baffling. For example, Waters is considered to be the least toxic, most reputable, and highest-ranked House member by X’s algorithm. This is despite her history of numerous falsehoods and toxic political and racial flamethrowing posted on her account.
- Radical Sen. Chuck Schumer (D-NY) is crowned King of Congress by X’s algorithm. While Waters is the highest-ranked House member, Schumer is the highest ranked among all 535 members of the House and Senate combined with a whopping 92 out of 100 “visibility” score on a 0-100 scale, indicating that his content is being boosted the most.
- X considers House Speaker Mike Johnson (R-LA) to be more toxic than anti-Israel extremist Rep. Ilhan Omar (D-MN). X’s algorithm considers House Speaker Mike Johnson’s (R-LA) account to be more toxic than that of Rep. Ilhan Omar (D-MN), who has a history of spewing “anti-Semitic” extremism. In June 2021, Omar put out a tweet equating the United States and Israel to the Taliban and Hamas terrorist organizations. She was also censured by the House in 2019 for her tweet accusing Jewish people of buying political support and mocking accordingly, “It’s all about the Benjamins baby.”
- X’s methodology is based on four metrics to produce a single “visibility” score for each account holder. The higher the visibility score, the better the tweets will perform. On a 0-100 scale, the average “visibility” score for Senate Republicans stood at an abysmal 66.8 out of 100, while Senate Democrats boasted a stellar 82.4 out of 100. In the House, the Republicans’ score was worse than the score assigned to their Senate counterparts, averaging a mere 58.6 out of 100. The House Democrats’ score, by contrast, was 67.0 out of 100, almost ten points higher than the House Republicans’ average.
STUDY: X Is Giving Congressional Dems a Sizable BOOST Before the 2024 Election
Introduction
Elon Musk has been outspoken about his intention to make X the preeminent digital town square and a platform free of censorship and political interference. But new evidence suggests a radical remnant within the ranks of Musk’s company is side-swiping his free-speech vision by resurrecting the infamous censorship techniques of Old Twitter.
MRC Free Speech America researchers discovered that X is actively boosting content posted by Democrat congressional members’ accounts, while simultaneously de-amplifying their GOP counterparts during an election year.
The X algorithm assigns each account a “visibility score” on a scale of 0-100. Each of these scores determines which accounts the platform boosts and which it de-boosts, according to X’s AI chatbot, Grok. MRC’s analysts prompted Grok to examine the social media platform's algorithm — which Musk open-sourced in 2023 in a display of transparency — to calculate the visibility scores for all members of Congress. The algorithm, per Grok, uses four core metrics, also based on a 0-100 scale, to calculate the overall visibility score. The results are in: X’s algorithm overwhelmingly favors the left.
Tilting the scales in favor of Democrats is well-trodden ground for the company, but it is now generally thought of as a thing of the past. These results suggest, however, a continued conflict within X about which policies should guide its operations.
On one side is Musk and his new approach to technology (like Grok and his commitment to algorithmic transparency); on the other side are legacy algorithms and certain holdover employees loyal to the prior censorship regime.
“I suspect the radical lefties at X are helping their radical besties in Congress, or it could be that the ghosts of Jack Dorsey are still plaguing the algorithm. Either way, it appears to be an inside job,” said MRC Vice President for Free Speech America Dan Schneider in response to our findings. He added, “It might make more sense if the moderates got treated the best by the algorithm, but they don’t. X assigns much better scores to the hard left than it assigns to moderates, and treats moderates better than conservatives. The X continuum consistently favors the left.”
Regression analysis of Grok-analyzed data confirms this trend. As rated by the Institute for Legislative Analysis (“ILA,” a non-partisan Congressional ratings organization), the firebrands in Congress with the most leftwing voting records are treated the best by X. Among these lawmakers are Reps. Maxine Waters (D-CA), Alexandria Ocasio-Cortez (D-NY), Ilhan Omar (D-MN) and Senate Majority Leader Chuck Schumer (D-NY), and Sen. Bernie Sanders (I-VT) (a self-proclaimed socialist who caucuses with Democrats).
X’s treatment of members of Congress generally gets worse as their voting records become more conservative. Based on their ILA-rated voting records, Sens. Tommy Tuberville (R-AL), Josh Hawley (R-MO) and Reps. Chip Roy (R-TX), Greg Steube (R-FL) and Scott Perry (R-PA) are all among the most conservative members, and their X visibility scores are among the worst. Whereas Sens. Shelley Capito (R-WV), Thom Tillis (R-NC), Lisa Murkowski (R-AK) and Reps. Don Bacon (R-NE), Thomas Kean (R-NJ) and Maria Salazar (R-FL) received relatively moderate ILA ratings, and the X visibility scores are worse than the most liberal but better than the most conservative.
One shocking result of MRC’s findings was that Waters is considered to be the least toxic, most reputable and highest-ranked House member by X’s algorithm with a total visibility score of 86.5 out of 100. This is despite her history of numerous falsehoods and toxic political and racial flamethrowing posted on her account.
The average score for Senate Republicans stood at an abysmal 66.8 out of 100, while Senate Democrats boasted a stellar 82.4 out of 100. In the House, the Republicans’ score was worse than the score assigned to their Senate counterparts, averaging a mere 58.6 out of 100. The House Democrats’ score, by contrast, was 67.0 out of 100, almost ten points higher than the House Republicans’ average.
In May, Musk assured the public that he is attempting to address the problem of reduced content visibility for certain users. Sen. Mike Lee (R-UT) asked Musk in a May 23 post, “[H]ow long will it take to get rid of the stage-five clingers at X—those who still periodically throttle conservatives?” Musk responded candidly, “Well, neither conservative nor progressives should be throttled. The point is to have an even playing field. I will investigate.”
Well, neither conservative nor progressives should be throttled. The point is to have an even playing field.
— Elon Musk (@elonmusk) May 24, 2024
I will investigate.
Schneider expressed hope, “Musk is clearly on their trail, trying to root out the problematic employees and algorithms in conflict with his vision. We hope that our findings will alert him to where these problems might be lurking.”
X’s Lefty Kingmaker Algorithm: How It’s Done
MRC researchers have used Grok to good effect. Musk’s AI platform spelled out exactly how the algorithm effectuates its political favoritism.
Grok revealed that X scored accounts across four separate categories, each on a scale 0-100.
The “Mass Appeal” category refers to an account’s diversity of followers. The “Reputation” category refers to an account’s alleged credibility. The “Toxicity” category refers to an account’s “offensiveness” or perceived “harmfulness.” Lastly, the “Follower” category references an account’s ability to retain followers. Each category is weighted differently to come up with a total final score that determines which accounts X boosts and which it deboosts.
MRC researchers prompted Grok with eight tests between July 1 and July 2 to assess how the platform weighted each of the metrics to determine a congressional member’s final visibility score. The four categories were weighted as follows: Mass Appeal (38 percent), Reputation (30 percent), Toxicity (21 percent), and Follow (11 percent).
Across all four categories, X on average consistently rated Democrat members of Congress better than members of the GOP.
X’s algorithm gave House Speaker Mike Johnson (R-LA) a high toxicity score of 42 points, which is worse than that of Rep. Ilhan Omar (D-MN) who had a 40 out of 100. The “Toxicity” metric is the only category where a higher toxicity is worse for the total visibility score. This means that the X algorithm considered Johnson more toxic than a notorious member of Congress who has a history of spewing “anti-Semitic” extremism.
In June 2021, Omar put out a tweet equating the United States and Israel to the Taliban and Hamas terrorist organizations. She was also censured by the House in 2019 for her tweet accusing Jewish people of buying political influence and mocking accordingly, “It’s all about the Benjamins baby.”
In another example of X’s glaring bias, the algorithm gave admitted-socialist Sen. Bernie Sanders (I-VT) an excellent rating of 82.4 out of 100, while libertarian Sen. Rand Paul (R-KY) barely managed a C-minus at 70.9 out of 100.
Congressional members's reputation scores reflected the most egregious examples of bias.
Grok noted that the X algorithm uses “Reputation” scores, which Grok dubbed as “Tweepcred,” to determine “the visibility and likelihood of a user's content being recommended to others.”
When MRC prompted Grok to explain further, it stated that “the minimum Tweepcred [reputation] score required for a user to be recommended on Twitter is 65,” indicating that any account with a lower score is de-amplified. Further, wrote Grok, “Accounts with higher Tweepcred scores are more likely to have their posts seen by a larger audience, while those with lower scores may have their posts deprioritized in the platform's algorithm.”
Javi Lopez, a tech expert and founder of image generator Magnify AI, wrote in an Oct. 25, 2023 X post that if one’s “Tweepcred” score “is above 65, you're in luck. Below that, you're out of the game. The shadow ban I experienced last week must have sent my reputation to the depths.”
The average reputation scores for Democrats and Republicans in both the House and Senate unveiled a glaring political bias: The average score for House Republicans was a paltry 55.7 out of 100 — meaning they’re severely limited in recommendation by the algorithm — while House Democrats enjoyed an above-65 average reputation score of 66.2 out of 100.
The bias was even more pronounced in the Senate, with GOP members scoring just above the 65 threshold with an average score of 67.6 out of 100. In contrast, Democrats boasted an astronomical 86.0 out of 100, giving the left an overwhelmingly undue advantage in terms of X recommending Democrats’ content.
X’s Algorithm Has a Bizarre Affinity for Senate Democrats Fomenting Dishonesty
X revealed its love affair with Senate Majority Leader Chuck Schumer (D-NY) and when it gave him the highest visibility score out of all of his colleagues in both the House and Senate. His left-wing colleagues in the Senate, by extension, also greatly benefited from this cozy relationship.
Out of 100 senators, the top 42 scores were exclusively bestowed upon Democrats. Republicans made up nearly all the bottom 43 ranked senators. Sen. Kyrsten Sinema (I-AZ), a former Democrat who split from the party, was the only exception.
Not only did X crown Schumer’s account with the highest overall “visibility” score (92 out of 100), it also rated him one of the most credible senators on the platform. X gave Schumer a stellar reputation score of 95 out of 100, appearing to overlook many major canards that the senator posted from his account.
Schumer’s X account has repeatedly pushed various fallacies. In one post, Schumer lent credence to former White House aide Cassidy Hutchinson, who gave what is now known to have been false testimony about Trump trying to wrestle the steering wheel of his motorcade from Secret Service agents.
Trump reportedly trying to overpower Secret Service and lead the insurrection is the lowest any president has ever gone
— Chuck Schumer (@SenSchumer) June 29, 2022
Presidents are supposed to protect Americans
Cassidy Hutchinson’s courageous testimony about the effort to undermine democracy should frighten every American
Schumer also posted the debunked claim that Trump had called white supremacists who attended the Charlottesville, Virginia Unite the Right rally in 2016 “very fine people.” A Snopes fact-check rated this claim “false.”
Nobody who meets with antisemites like Nick Fuentes, called white supremacists in Charlottesville “very fine people,” disgustingly said Hitler “did some good things” has any right to lecture Jewish Americans on their personal political beliefs
— Chuck Schumer (@SenSchumer) March 19, 2024
Everyone must condemn Trump’s words pic.twitter.com/EJgRLIr1mC
But Schumer isn’t the only Senate Democrat that X’s algorithm highly regards.
Sen. Elizabeth Warren (D-MA) is one of the highest-ranked members of Congress by X’s algorithm. X gave Warren’s account the lowest toxicity score in the Senate despite her oftentimes vitriolic language.
Warren used her account in 2017 to smear Trump’s then-Attorney General Jeff Sessions as a racist and a sexist and even outright threatened him.
If Jeff Sessions makes even the tiniest attempt to bring his racism, sexism & bigotry into @TheJusticeDept, he'll hear from all of us.
— Elizabeth Warren (@SenWarren) February 9, 2017
X also granted Warren’s account the highest reputation score of any senator, but her post history tells a different story. One Warren post falsely blamed the unfortunate suicide of an overdose victim as a “killing” perpetrated by anti-trans “extremists.”
She actually killed herself.
— Andy Ngô 🏳️🌈 (@MrAndyNgo) March 13, 2024
Source: The Oklahoma Chief Medical Examiner
Meanwhile, X’s algorithm punished and downranked Republicans, including those with a history of being on the right side of the truth.
Consider Sen. Rand Paul (R-KY). X put Paul in the bottom half of the Senate in overall “visibility” scores (70.9 out of 100). In the aftermath of the COVID-19 pandemic, Paul, who is also an ophthalmologist, emerged as one of the preeminent voices on the origins of the virus on the platform. Paul consistently argued that COVID-19 was likely a man-made pathogen created in the Wuhan Institute of Virology. President Joe Biden’s Federal Bureau of Investigation and his Department of Energy have both since affirmed that COVID likely came from a lab.
X also significantly de-boosted the account of Sen. Eric Schmitt (R-MO), ranking him in the bottom five in the Senate with a failing 59.4 out of 100 visibility score. Schmitt has continuously called out Biden’s White House for verifiably censoring Americans even while many on the left derided the allegation as a baseless “conspiracy theory.” Schmitt’s account also suffers from a low 55 out of 100 reputation score, below the 65 “Tweepcred” threshold needed to be recommended by the algorithm.
X’s bias barely abated even when calculating the average visibility scores just for senators running for reelection in 2024.
When reviewing the individual account scores of Republican and Democrat Senators running for reelection, X’s algorithm displayed a 14-point preferential bias in favor of Senate Democrats.
Republican candidates received an average score of 68.6 out of 100 while Democrat candidates received an average score of 82.5 out of 100.
X’s algorithmic favoritism threatens to subvert the outcomes of some of the most highly competitive senate elections in November, as social media platforms such as X provide candidates with indispensable resources such as donations and exposure to on-the-fence voters.
X Continues Its Inexplicable Favoritism in the House
While only one-third of Senators are running for reelection this year, nearly every House member will be vying to retain control of their seat in November. The X algorithm is seemingly playing a similar censorship game in the House, which is ultimately benefiting the left and diminishing the reach of House Republicans on the platform.
X gave far-left House Democrats a distinct advantage over their GOP opponents, just like it did for members of the Senate.
House Republicans received an especially low average score across all four “visibility” metrics with a 58.6 out of 100. Conversely, House Democrats scored almost ten points higher with an average of 67.0 out of 100.
Inexplicably, Rep. Maxine Waters (D-CA) received the highest visibility scores out of any member of the House. X’s algorithm assigned her an overall score of 86.6 out of 100.
X’s algorithm ignored Waters’s habitual tendency to purport false and toxic narratives, as the social media platform rated Waters with a reputation score of 87 out of 100 and a meager toxicity score of 9 out of 100.
This bafflingly high score seems to ignore Waters’s repeated trend of posting false narratives on X. After the high-profile case of Jacob Blake’s fatal encounter with police, Waters proceeded to smear the later-vindicated officer involved in the shooting on X (then Twitter) as a potential murderer without evidence.
Another unarmed black man, shot in the back, by a WI police officer! No matter the outrage at the killing of George Floyd by a police officer, knee on his neck, the killing of black men continues. Is this a defiance by the police that indicates they don’t intend to stop?
— Maxine Waters (@RepMaxineWaters) August 26, 2020
Waters also had a chronic habit of launching Twitter tirades against Trump. She disparaged Trump as a “mad man,” and a “deranged lunatic.”
The president is a deranged lunatic & he’s inspired white supremacists to violence, the latest of which was a plot to kidnap Gov. Whitmer. These groups have attempted to KILL many of us in recent years. They are following Trump’s lead. Dems & GOP must agree to push for 25th Amdt!
— Maxine Waters (@RepMaxineWaters) October 9, 2020
Rep. Adam Schiff (D-CA) is another disreputable Democrat that X’s algorithm considers trustworthy. Schiff was a vocal peddler of the debunked Russia-Trump collusion conspiracy, which was revealed to be a coordinated campaign perpetuated by the Hillary Clinton campaign in 2016 and members of the top law enforcement and intelligence agencies to smear Trump as a Russian puppet. Schiff, who was the head of the Intelligence Committee, famously claimed that he had “ample evidence” of Trump working with Russia. He tweeted false claims numerous times about Trump as well as the Mueller Report which cleared Trump of those allegations.
Robert Mueller was clear:
— Adam Schiff (@RepAdamSchiff) July 28, 2019
Russia interfered in our election to help Trump. Trump's campaign welcomed and used this illicit help.
They never reported it to the FBI. Instead, they lied about it to investigators.
And used those lies to obstruct justice. pic.twitter.com/EVfYRrfn8p
Conclusion
Musk’s $43 billion acquisition of Twitter in 2022 was a major turning point in the free speech fight against Big Tech censorship. But it’s becoming apparent that holdovers from Old Twitter aren’t letting go of the power they tasted during the 2020 election.
The theory of an internal revolt among Old Regime holdovers at X is also not a new one.
Using its CensorTrack.org database, MRC revealed a shocking spike in censorship cases in 2023 following Musk’s acquisition. This was a major red flag showing how Old Twitter employees had actively undermined Musk’s pragmatic approach to free speech online.
This time, however, Musk’s AI and his actions to make the algorithm transparent are making it much harder for the radical remnant to hide their insubordination from Musk.
METHODOLOGY: Between June 7 - June 25, MRC researchers gave Grok specific prompts in order to ascertain “visibility” scores for individual members of Congress. The questions were specifically tailored so Grok would only report the essential criteria that would determine whether or not a specific account on X is pushed/recommended to other users on the platform (“Mass Appeal,” “Reputation,” “Toxicity,” and “Follow”). Grok calculated numerical value scores between 1 and 100 for each member of Congress. MRC researchers then calculated the average scores for all members of Congress across all four criteria used by the algorithm to determine X’s “visibility” scoring process. To determine how X weighted each of the four categories in determining a “visibility” score, MRC researchers prompted Grok with eight tests between July 1 and July 2 to assess how each of the metrics were weighted to uncover which was considered most important when determining whether or not an account would be boosted. Grok repeatedly provided information with minor variations on how each category was weighted to determine a congressional member’s final “visibility” score. MRC researchers then took the average weighted percentages across the eight tests to deduce a mathematical formula. When the averages were computed, the categories were weighted as follows: Mass Appeal (38 percent), Reputation (30 percent), Toxicity (21 percent), and Follow (11 percent). “Mass Appeal” and “Reputation” scores were consistently in the top two spots in measures of importance across the eight tests.
