Berlin. Chancellor Friedrich Merz (CDU) is only occasionally on social media platforms, he revealed to the magazine “Der Spiegel” at the end of April. “If you look at what is being spread about me there, how I am attacked and degraded,” said Merz.
Far more than one in ten comments about Merz contained conspiracy stories or calls for violence against him. According to the online evaluation, bots and influence from Russia also shaped the election campaign in Saxony-Anhalt, Mecklenburg-Western Pomerania and Berlin.
Penemue analyzed more than 120,000 comments on the platforms Tiktok, Facebook, Instagram and
Penemue has trained this over the past few years, says co-founder and board member Jonas Navid Mehrabanian Al-Nemri – “mainly for recognition in texts, but also in combination with emojis, coded language or algospeak”.
By this he means, for example, codes and symbols whose meaning only becomes clear from the context or within a certain group - for example a blue heart as a symbol for AfD sympathizers. Part of the analysis is always human validation of the content. However, there remains a residual uncertainty; there simply cannot be 100% certainty with such interpretations.
Penemue classifies all hate comments. The company considers offensive statements to be “toxic”. It defines those that are in the criminal area as “harmful”. These include anti-Semitism, xenophobia and threats of violence. “These are misanthropic statements, often group-related misanthropy,” says board member and co-founder Sara Egetemeyr.
Although the Chancellor did not run in any of the three elections, Penemue recorded more than 3,000 Merz-related comments for the election in Saxony-Anhalt - 93.1 percent classified them as negative or "toxic", 22.9 percent of them as "harmful".
In the analysis of Berlin and Mecklenburg-Western Pomerania, 84.6 percent of the 996 comments about the Chancellor were toxic, of which 13.2 were harmful. What is particularly striking is that not even one percent of the comments about Merz examined were positive.
Nevertheless, Merz did not receive the most hate comments. “Manuela Schwesig absolutely has the highest volume of comments,” says Egetemeyr about the SPD’s top candidate in Mecklenburg-Western Pomerania. Schwesig received 19,069 comments. A little more than half of them are “toxic”, almost 16 percent of them are “harmful” – but almost a quarter of them are also positive.
In addition to insults, the harmful comments also included a high proportion of xenophobic and Islamophobic statements. Elif Eralp (Left Party/Berlin) and Claudia Müller (Green Party/Mecklenburg-Western Pomerania) were particularly affected by misogynistic statements. Politically active women, politicians and activists in particular experience sexism and misogyny online, according to a 2025 study by the Bavarian Academy of Sciences.
Disinformation and conspiracy stories also characterized the comment columns examined - both in articles about Merz and Schwesig and in the election debate as a whole.
In addition to the many answers from real people, there were also some comments from suspected bots. These are automated accounts that are active on the Internet. They often spread viral advertising in comments, but sometimes they are also used specifically for hate speech.
Penemue examines their emergence through several factors, looking at frequency. "Does an account post on several threads at the same time every second, perhaps even always at the same intervals? A human can never reproduce that," says Mehrabanian Al-Nemri.
“We also found comments with the same words very often: the same sentence is repeated everywhere in a spam-like manner,” he says. At Schwesig and the state SPD, the company assigns around a tenth of the comments to possible bot activity.
However, the classification must be made carefully, says Lea Frühwirth, disinformation expert at the Berlin Center for Monitoring, Analysis and Strategy (CeMAS). “The label ‘bot activity’ is always an attribution from outside because we don’t have any data internal to the platform,” she says. “Certain behavior patterns make it more likely that you are dealing with a bot. But there are also outliers of human behavior that are unusual but still authentic.”
CeMAS investigates disinformation campaigns by state actors such as Russia, for example “Matryoshka” or “Storm-1516”. Over the summer, these campaigns were responsible for the spread of fake news videos. Merz and the Handelsblatt became their targets. The Office for the Protection of the Constitution warned of increased disinformation activity around the state elections.
The Kremlin uses bots, among other things, to spread false information and fuel hatred online. In the election in Saxony-Anhalt, Penemue attributed around 7.7 percent of the comments to “potentially Russia”. However, this cannot be said exactly because the IP address can be easily disguised with a VPN.
Clues could include time zones or translation errors, such as phrases translated verbatim from Russian that don't make sense in German. But: “Not every comment that is pro-Russian fits the pattern of the well-known disinformation campaigns,” says Frühwirt. It is not known that the comments evaluated by Penemue are part of the known campaigns.
Although it remains unclear who is behind suspected bot comments, their mechanism of action has been investigated. Egetemayr says: “They rather set selective accents in order to then change the discourse.” A strategically placed provocative comment - such as a xenophobic statement under a post from the left-wing party - could attract users, who in turn write further comments about it.
The extent to which these online mechanisms affect German voting behavior remains speculation. “It cannot be methodologically measured whether the amount of disinformation led to a specific election result,” says Frühwirt. Those who are exposed to false information more often are influenced more than by a single post. “But the decision at the ballot box depends on so many different factors that then create an opinion,” she says.
