Did Bots and Right Wing Media Platforms Influence How Germans Voted in the February 2025 National Election? Part 1

Evidence shows that X favored the AfD. Does this prove that the German elections were influenced by X?

Europe Calling is a pro democracy political policy organization. Their website says,

Europe Calling is probably the largest policy webinar format in Europe. Since 2017, more than 250,000 people from all over Europe and the world have regularly participated online, engaging directly with policymakers, leading academics, stakeholders and civil society representatives.

As a community, we strive to create an open transnational space for European citizens to engage, learn and organise a voice in the key European debates of our time.

Our goals:

• to make political processes, their backgrounds and political and economic power structures transparent and thus accountable;

• to strengthen the democratic civic society beyond national borders in Europe and internationally;

• to strengthen the quality of political decisions through citizen participation;

• counteracting disinformation, fake news and political populism through a culture of participation and democratic discourse;

• to open up access to interested citizens for democratic civil society organisations and their issues;

• to demonstrate opportunities for influence and participation and to support citizens in making use of these opportunities.

So far I have attended two of their webinars. I am going to share the research from Europe Calling “Free opinions? – How X and TikTok influenced the German federal election” According to a comment the host Maximillian Fries made, there were about 2,500 people in attendance. It was in German and English with the option to have a translation into either of the languages, as well as sign language. Most of the discussion was conducted in English. The two presenters were Computer Science Professors Prezmyslaw Grabowicz and Karine Caunes. There was also a German Member of European Parliament Alexandra Geese. She is a member of the Green party and a digital expert. In Part 1 I will discuss the research of Prof. Grabowicz.

Prof. Przemek Grabowicz started off with a video where he showed Elon Musk speaking on the X platform and telling us that people can’t make the right decisions if freedom of expression is not allowed. They say that Elon Musk sees himself and his platform as champions. [Do you think he even believes this stuff?] Then it showsUS Vice President JD Vance and calls him Musks loudest supporter. They say, “the Trump administration and the Tech bosses are fighting together against the EUs digital laws for freedom of expression as Musk likes it. But how far does freedom of expression go on X? We asked the users.” Then they interviewed two young men. One said he thinks freedom of expression is important, the other said he does not think hate speech is just opinion not fact, and his tone implies that since it is not factual it is not necessary.

Next, they lay out the problem. Prof. Grabowicz wants to see, “Does X keep its promise to stay neutral?”

The narrator tells us that, “They use two dolls to represent 2 new accounts. They follow 64 politicians. Eight from each party in the Bundestag, but they don’t read, like or comment on anything, so for X’s algorithms they are users without any traits.” I see that somewhat differently. Since, they are following politicians from 8 parties, that would not be typical, and might imply something about an account where the user does that. There may be an algorithm for those people. The video also does not tell us is whether these users have gendered names. I have heard that the “normal” algorithm for Facebook for example is a White, male, Christian. So, is X assuming this too? We don’t know. The study also collected data on all German politicians who are on X.

Throughout January and February of 2025, Prof. Grabowicz helped his team of graduate students in Dublin and at Amherst who followed the social media accounts with their data analysis. He says that they checked the newsfeed on the accounts every 30 minutes.

According to the video narrator, X offers two feeds, a “for you” feed and a “following” feed. “The first is the user selection. The second is Xs recommendation.”

They analyzed the data. First they looked at how often each party posted something to read. They saw that the centrist parties the Greens, and the FDP posted more than the AfD. (see the chart below)

Then they looked at what X was offering their two accounts to read and they found that AfD and BSW content was offered more than twice as often as other parties. (See the chart below.) You can also see how little X offers SPD content. In fact for the SPD, Greens, FDP and CDU they are represented in offerings much less often than they post whereas the other parties are represented more often than they post. It looks like the AfD and BSW are being amplified by X.

The study asks the question, “why are feeds from the political fringes shown more often?” They see that it is clearly not from interaction alone. They confronted X with the findings at their EU headquarters in Dublin and got no response!

The narrator goes on to tell us that Elon Musk interfered personally in the German election campaign. I assume they refer to Musk interviewing AfD chancellor candidate Alice Weidel on X to promote her, and his saying that Germany needs to vote for the AfD. That begs the question of how neutral his platform is. In January X recommends one candidate more often to the test accounts, the AfD chancellor candidate Alice Weidel. Even though the accounts don’t follow Elon Musk they also are shown Elon Musk’s tweets the most often, such as the one below.

We hear that Alice Weidel’s tweets reached 15 million users compared to the usual 1 million for the other candidates. That is around one quarter of the 59.2 million eligible German voters. The candidates for the Greens (Habeck), CDU (Merz) and SPD (Scholz) did not get a boost from X. The study asks, “Is that still free speech?” It seems to me from Musk’s point of view it is free speech for me, and not for thee.

They analyzed the pro AfD tweets. Elon Musk appeared most often followed by right wing US influencer Alex Jones. Secondarily we see two AfD party members as biggest tweeters. Dennis Hohloch is from Brandenburg where the AfD received 29.5 percent of the vote. It came in second just behind the SPD with 31%. Stephan Brandner is from Thüringen also a former East German state, where his AfD party won first place with 30.5% of the votes.

We learn that in 2022 Thierry Breton, EU Commissioner for the Internal Market traveled to Elon Musk to explain about a New EU law about online platforms. Elon Musk supposedly agreed to comply with it. That would include giving requested interviews and data. The new Digital Services Act (DSA) was supported by 90% of the EU Members of Parliament (MEPs). “The act requires platforms to take measures against any actual or foreseeable negative impact on Public Debate, Electoral Processes and Public Safety.” Elon agreed, but then when confronted, his company did not comply.

In February Prof. Grabowicz and his team shared the data of their study with the EU commissioner from Finland Hanna Virkkunen who has been investigating whether X violated the DSA law since January 2024. She says the EU has to defend the shared values, and they would never use this digital law as a bargaining chip with the US. However, I have recently heard that a fine that has rightfully been levied on X will be made clearer this summer. Perhaps that is why Elon Musk is promoting the idea that the USA should have zero tariffs with the EU. He has a Tesla factory in Germany, his and his sales are dropping drastically in the EU. The EU is also seeking to become independent from Starlink. So Musk might want to come across as someone who supports the EUs financial interests. Will this work?

The study looked at the reasons that might influence the number of times a party would be appearing in the feeds. I was a little unclear what was meant by Number of Engagements. Prof. Grabowicz’s explanation was that it was people engaging with a feed. So, that would be the behavior of people other than their two dummy accounts who posted to or liked certain material, and how often the politicians being followed engaged with X. He said Alice Weidel does this quite a lot. His dummy accounts did not have political affiliation a factor that influences feeds, and since they were also not posting they would not be followed for that reason.

Given the nature of the study the accounts have no personalities, so, no interests, no posts, and no private data. One could assume that X would not know which trending topics would interest them, except for the ones having to do with the 8 politicians they followed.

Again, Prof. Grabowicz says what they found was that the number of engagements by people most explains their feeds. However, since the accounts were not engaging other than to follow some politicians it would have to be something else explaining their feeds. The study shows that the political affiliation of the people posting a tweet seems to matter. Grabowicz feels the perfect model would explain 100% of the variance in the feeds, so with around 35% of the variance which he feels is getting high given that they are missing necessary factors to explain their results. It seems low when one sees 100% as the target.

The DSA allowed Prof. Grabowicz and his team to do this research. The law relies on people like Professors Grabowicz and Caunes to study platforms like X to make sure that the effects are appropriate. Without being able to talk with people at X Grabowicz’s team is missing factors in the models which would allow them to fully explain what is going on. Even though the law says that social media platforms have to comply with these requests X did not. The research shows that X seems to promote the AfD, but the researches felt that there is no proof from their study that shows the increased exposure to the AfD caused people to vote for them. Grabowicz told us that since the AfD’s percentage of support only went up by 1-2% in the time period of the study it might not have influenced them at all. At the same time as an observer of their data one can see that the 1-2% could be cultivated by being bombarded with messages on X. We do not know what happens to real users from this study but can only extrapolate that with more knowledge of the real users X was making their messages more targeted to the personal preferences. It certainly seems like something we should be analyzing in US elections too, but the US does not have this sort of Digital protection law.

Grabowicz ends by asking the question of whether Elon Musk has the right to influence elections. He points out that Elon Musk is a foreigner trying to influence an election that is of another country, and therefore he does not have the right to influence the election, not legally. Not in the EU.

I will continue with the research by Professor Karine Caunes in Part 2.

https://arstechnica.com/tech-policy/2025/04/eu-may-make-an-example-of-x-by-issuing-1-billion-fine-to-musks-social-network/

https://carnegieendowment.org/europe/strategic-europe/2025/02/russian-interference-coming-soon-to-an-election-near-you?lang=en

https://europe-calling.de/en/webinar/european-spring/

https://www.destatis.de/DE/Presse/Pressemitteilungen/2024/12/PD24_460_14.html

https://www.theguardian.com/world/2024/sep/23/germany-social-democrats-sdp-brandenburg-election-aftermath