In January 2024, more than a million people took to the streets in Germany. They were appalled by the right-wing party AfD’s apparent plans to deport migrants and minorities after German news outlet Correctiv published an investigation into a secret far-right meeting in Potsdam.

As shocking as those revelations were, they were not surprising. Nor were they new.

For the past 10 years, the AfD — by now the second-strongest party in Germany’s national parliament — has been shaping such narratives. Not just behind closed doors but also in the eye of the public and at the heart of democratic discourse: in Germany’s state parliaments.

This is what our team at the Tagesspiegel Innovation Lab, in collaboration with researchers from Freie Universität Berlin, set out to show.

With the help of language models, we analysed the full corpus of speeches delivered by AfD representatives in all 16 state parliaments since the year the AfD gained its first seats in state parliaments in 2014, encompassing over 15 million words from more than 340,000 speech paragraphs.

Our data analysis revealed that concepts such as “remigration” — now in the national spotlight — have been part of the AfD’s political vocabulary for years.

Taking a closer look

We — an interdisciplinary team of journalists, programmers and designers, in collaboration with political scientists — asked ourselves: What exactly has been said in state parliaments? What do we learn when we stop looking at individual sound bites and start looking at the full dataset? And how can we effectively visualise this slow, steady shift in rhetoric?

The foundation of this project was the new StateParl dataset, which provides systematic and automatable access to German parliamentary transcripts for the first time.

The team used a language model to detect xenophobic content within this vast dataset. With a combination of qualitative analysis and machine learning, we identified recurring patterns — ways of speaking that can shape narratives and alter a discourse over time.

What emerged were four overarching storylines that have dominated AfD rhetoric in Germany’s state parliaments:

• The criminal foreigner: A consistent linking of migration with violent crime, particularly knife attacks, despite a lack of statistical evidence.
• Cultural incompatibility: The portrayal of immigrants as fundamentally unable to integrate, with a rhetoric that casts mosques, for example, as symbols of foreign conquest.
• The “Great Replacement” theory: A conspiracy narrative suggesting migration is part of an orchestrated effort to replace the native population, echoing ideas found in far-right circles across Europe and the U.S.
• “Remigration” as a political goal: The repeated call for the mass deportation of people perceived as not belonging, presented as a policy proposal.

These narratives do not occur in isolation. They build on each other. They reinforce one another.

Covering hate speech without encouraging it

But covering hate speech poses an ethical dilemma: how do we report on dangerous rhetoric without amplifying it? Our approach was to stay rooted in data and contextual framing. Direct quotes from parliamentary debates are shown selectively, embedded in a broader analysis that traces their origin, evolution, and societal effect. By using our interactive tool that accompanies the article, readers can search the database for keywords themselves.

This effort sheds light on how xenophobic ideas gradually become part of everyday political discourse. As such rhetoric becomes more common in official forums, it reshapes norms, making once-unthinkable positions seem debatable.

Parliaments carry legitimacy. What happens in them shapes public debate. The AfD has used this space effectively to shift what is acceptable to say — and, by extension, what seems possible to do.

The idea that extremist rhetoric fades once a party enters formal institutions is a comforting one. But our data suggests the opposite: the longer the AfD is present in parliament, the more radical and confident its language becomes.

This is why this investigation is not just about numbers or keywords. It’s about understanding how democratic spaces can be utilised to introduce anti-democratic ideas. It’s about recognising the slow, deliberate process by which political norms are stretched — and sometimes broken.

Banner illustration by Markus Günther.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .