Did Bots and Right Wing Media Platforms Influence How Germans Voted in the February 2025 National Election? Part 2

What do X and TikTok have in common in influencing the German elections?

Apr 18, 2025

French Law Professor Karine Caunes works on making sure that the EU Digital Services Act can be used to avoid manipulation of people by social media. She is also working on a board that is developing a code of conduct for AI models to help make AI safe. Her research asks the questions we see posted above. “Is there freedom of expression and information online? If there are threats, where are they coming from?” She applies these questions to examining the effects of X and TikTok on the German elections held in February of 2025.

She conducted 2 studies. One was on the influence of X on the German elections they did 2 studies. The other was the influence of TikTok on the same elections. Because the social media are different the perimeters of the studies were different and the timelines were not exactly the same so they do not have an exact comparison.

For the study of X what she analyzed was all of the tweets and retweets of any candidate from a main German party from January 4 to January 16 2025. She chose this period because she knows that when you have astroturfing, which is the manipulation of elections with the help of fake accounts called Bots, it usually happens within 2 months before the targeted election. It was also a period when Trump was not yet in power and Elon Musk was not either. Still, Musk already had a huge influence on Trump before Trump’s inauguration on January 20, 2025. She looked at tweets from, to and citing @ the different parties and analyzed almost 500,000 tweets issued by more than 130,000 X accounts. In the image above the blue area is the pro-AfD community. It is bigger than all the others. This was just the political parties’ tweets not a fan of the political parties tweeting to another fan.

What they found in analysis is that most of the political parties had very organic interactions, with discussion and comments within their communities. By organic she means humans communicating with each other. However, the AfD had a huge influx of posts from outside of Germany, and that was from Elon Musk. One can see his name in an area of tweets on the previous slide. In the AfD posts there were mostly tweets and retweets and very little discussion. Discussion is generally considered to be organically generated. Instead, the behavior of the accounts with just tweets and retweets with very little comment is the behavior one sees with Bot accounts.

When they examined whether there were many fake accounts created to market a campaign called astroturfing they found that there were. One way they could tell this is that they had thousands of accounts that were tweeting every few minutes around the clock, months on end, something no human could do. Most of these accounts supported the AfD. Then, right before Elon Musk was going to interview Alice Weidel there were suddenly more than 2000 new Bot accounts created in 2 days, that promoted the interview. The new accounts only followed accounts supported by Elon Musk, so one could think that he is behind their creation, especially since it is on his platform. Prof. Kaunes says that we can see that a large part of Elon Musk’s following is Bot accounts. I would love to know what percentage of his following is not real

She goes on to tell us why they look at Bot accounts. If you are looking for freedom of expression, Bots cannot do that because they are not human. If they predominate in a social media by tweeting and retweeting, they give priority to some content and not others. That has to do with the algorithms that social media use, so in a sense one can use bots to manipulate the algorithms, or rather, to take advantage of them. So they can suppress the freedom of speech of some people, because they engage with the content of some and not others. In their study they saw that this artificial inflation of content was in favor of the AfD.

In the graph on the slide above you can see the distribution of bot accounts, and can see that the AfD has the most, while Die Linke has the least. Interestingly enough, Die Linke also doubled the number of people voting for them, but they apparently did this with mainly organic and therefore legitimate engagement. More than 95% of the Bot accounts support the AfD. Professor Kaunes also points out for example that there is a Bot called Biby which only retweets content from 5 accounts. Alice Weidel and 4 other members of the AfD.

She illustrates the difference between Bot behavior and normal person behavior. The average person has 19 tweets or retweets a day, the top 1 percent of accounts average 321 tweets or retweets a day, since their creation. This non-stop tweeting and retweeting is not humanly possible, so one assumes they are Bot accounts.

We see examples of accounts that have AI generated content in German, and because they have Blue Marks by the names one could think they are legitimate accounts, because that mark means X approved the accounts, but they are really fake accounts. She says it is a typical strategy for tricking people into thinking they are interacting with real people. She also points out that there were 364 accounts talking about the German election for the year prior to January 2025, but suddenly in January of 2025 there were over 5000 accounts created in 17 days. This wave usually indicates that there is an astroturfing campaign and most of them support Musk and the AfD community.

Professor Caunes concludes that Elon Musk provided the AfD with impact that was greater than others by tweeting and retweeting material in support of them. While some of these pro AfD Bot accounts generated AI responses to people, it was detectable that they were Bot accounts because of the number of tweets and retweets that they engaged in from the time of their creation. The sheer number of these accounts boosted the AfD visibility. Professor Caunes compares this with data that the Bundesdatenschau developed as well.

Here we can see that they looked at the X affiliation of all members of Parliament. If you look at their data before the election period, shown in the first graph, the Green party had the most visibility meaning the number of times someone looked at their tweets, and the AfD was in third place. However, the second graph shows the number of times people are looking at tweets during a 2 week period in January 2025 and the AfD is way ahead of the other parties. The number of times people are looking at a parties’ tweets is not correlated with the number of times that people in that party are posting. We see that instead it correlates with the number of tweets and retweets which have mostly been produced by Bots. Prof. Caunes calls this inflation of one party’s viewing and subsequent repression of others through using Bot accounts, “reverse censorship.”

Professor Caunes second study was of TikTok. She did not look at political parties themselves, but at hashtags which referred to political parties or names of leaders of political parties. That means there could be comments from anyone on TikTok. This is how the Romanian elections had been analyzed on TikTok, and she wanted to see if one could see similar trends in the German election. They analyzed more then 2,000 channels and more than 1 million commenting accounts linked to almost 3 million comments. You can see that there are 2 major communities in the image above. The colors do not correspond to the political parties, they are random. In this one the red community, which is the biggest one, represents those comments that are pro-AfD. The blue represents comments pro-Die Linke, which is the second biggest community.

She checked for Bot influence campaigns and found that they had one with the AfD, but not with Die Linke. She points out that in the German election when you look at the youth vote they voted primarily for the AfD, and secondly for Die Linke. She says there may not be a causality, but there is a correlation with the amount of comments.

Professor Caunes interpreted the data as showing that the playing field is not level amongst the candidates, because Alice Weidel is overwhelmingly supported by TikTok accounts. If a person looks at some content and has a 50% percent chance of being a fan of Alice Weidel because that is who viewed the same content a lot, then the next content the person will be shown is some supporting Weidel. If the person looks at other content that supports some other party such as Bundnis Sarah Wagenknecht, they will also be shown content of Weidel afterwards. The algorithm appears set up to support certain viewers in seeing pro-AfD content.

An example she used to illustrate how this works is that if Sarah Wagenknecht makes a comment, then you would see a lot of blue hearts on a comment. The blue hearts typically represent the AfD. That means that the next content the person will see is content from the AfD. This is spam content and it is very influential in the TikTok recommender system. Many accounts were targeted with these blue heart spams. Not just team Sarah Wagenknecht, which makes sense to me, because she sounds a lot like AfD, but also TeamOlafScholz, and some journalists were targeted with this AfD promoting spam.

They also looked at illegal content on TikTok. They found that TikTok never removed illegal content such as German history revisionism or immigration discrimination. These would both be areas of content that supports negative points of view of the AfD.

Professor Kaunes ends by saying telling us to her question of “does TikTok support freedom of expression?” that it only does if you think Bots are okay. However, she seems to think that Bot accounts should be removed from social media platforms so that people can interact only with other human beings. Bots clearly are attempting to manipulate people’s thoughts, but that does not prove that they influenced their votes. This cannot be determined by her studies with the information she had available.

Caunes points out that according to the Digital Services Act all social media should be doing risk assessments of risk to political discourse and freedom of expression, freedom of information and freedom of thought. She says that all she is asking is that the European Commission clean up the internet from Bots and illegal content. That does not bode well for X or TikTok if there is to be accountability for these social media platforms from these studies, given what we have seen. She points out that X is legally required to explain their algorithms. They could look at the data, since X was required to keep it and then they could look at the algorithm and determine what is going on.

While her studies cannot directly prove that people were influenced we can see that there were attempts to manipulate people by the existence of the bot accounts supporting AfD. Professor Caunes sees that TikTok, which we view as having a youth audience does promote AfD and Die Linke, which she sees evidenced in these two parties getting the most youth votes in proportion to their community support levels on TikTok. She also proves that there are at least 2 influences on social media which is that there are Bot accounts, and there is illegal content supporting AfD values which is not being removed on TikTok.

We know that these two social media platforms are both foreign owned, and they did amplify the AfD. We can see this action as foreigners interfering in German elections. This is illegal. The consequences are that the companies can be fined for up to 6% of their global worth. For Elon Musk that would mean all of his companies combined. We know that with the current trade tariff wars, allowing these social media platforms to continue to operate while violating EU laws is something that Trump is trying to negotiate. I hope that the EU leadership does not negotiate away control over the usage of social media, and that they do fine these companies as they are supposed to. It would set a dangerous precedent if the EU does not follow their own laws. Allowing other countries to manipulate your elections, be it with social media platforms or with money or with both is the slippery slope that all the monetary trade advantages cannot make up for as you lose grasp of the country and society by allowing your citizens to be manipulated by bad foreign actors. At the same time one should hold the political parties who benefit from this accountable too. I think that the AfD should be fined as well.

Linda on Life Abroad

Discussion about this post