Method

What this site does

parteiprogramm.com summarizes each party's election program point by point. Every point is written in plain language and in an even simpler version with short sentences and common words. Each one comes with a word-for-word quote from the program and its page number, so you can check every statement yourself.

The site shows every party that has seats in the Bundestag, in a state parliament or in the European Parliament and has published a current election program for the federal level or a state. We do not place programs for European elections. For a party whose only seats are in the European Parliament, we show its program for the federal election.

The site is independent and has no connection to any party. Who runs it and why it exists is on the About the project page.

The content is written by AI

An AI reads the program, splits it into individual points, writes the summaries and explanations and places each point on the axes below. For every placement it writes one sentence of reasoning. Every program goes through the same instructions.

No person checks each individual point before it appears. Automated checks make sure that every quote stands word for word on the stated page, that our texts take over no longer runs of words from the program, and that the simple language uses short sentences and words. Nobody checks whether a summary renders the content correctly. Summaries, explanations and placements can therefore contain mistakes. The party's program is always what counts. Every point links to the right page of the original.

Quotes back up the summary

Every point comes with a short quote, word for word from the program. On the topic pages, you open it under the point. Below it is the page number with a link to that page of the original. So you can see for yourself what our summary rests on.

Before a quote appears, we check automatically that it really is on that page. Below each quote is our own English translation. The German wording is what counts.

The axes

Each axis has two poles. A point is placed only on the axes it actually touches. A demand about pensions usually says something about the economy but nothing about migration. The migration axis then stays empty for that point.

The economy axis runs from "more state" (more redistribution, more public services, more rules) to "more market" (more competition, more personal responsibility, lower taxes). We do not use the words "left" and "right", because in everyday speech they mean many things at once. The foreign policy axis covers foreign, defence and alliance policy, such as the armed forces, NATO, support for Ukraine, arms exports and development aid. No axis says whether a position is good or bad.

The scale

Every placement is one of five steps. The yardstick is the law at the time of the program: we measure how strongly a point wants to change the current state. "Leaning" means it shifts it step by step. "Clearly" means it changes it at the root, for example by abolishing something or introducing something entirely new. Here on the economy axis as an example:

How the values come about

For a whole program, the value of an axis is the average of all points that touch this axis. Points without a placement on an axis do not count there. Next to each value you see how many points it comes from. A value from three points says less than one from thirty.

Under each bar, small columns show how many points sit on each of the five steps. It shows whether a program is moderate throughout or has strong demands next to moderate ones.

As a check, we compute each value in two more ways: with the point in the middle, once all points are sorted, instead of the average, and with an average in which every topic counts the same. If one of these values differs from the average by at least one whole step, the value carries the note "Depends strongly on the method".

The compass shows the economy axis horizontally and the society axis vertically, with "conservative" at the top. Next to it, all seven axes appear as bars. Each topic page applies the same calculation to that topic's points only.

On the home page, each party appears with its main program. That is its current program for the federal election. Parties without a federal program, such as regional parties, appear with their state program. CDU and CSU have a joint program for the federal election, which appears for both parties.

What we follow and how we check

We take two things from political science.

• Place single statements, then average. This is how the Manifesto Project works, which has analysed election programs for decades: it splits a program into single statements, codes each one and computes the party's position from them. Here an AI places every point, and the program's value is the average. Le Mens and Gallego examined in 2025 whether language models can place political texts in this way.
• Look at economy and society separately. Research does not describe parties on one line from left to right, but in at least two dimensions, an economic and a social one. The Chapel Hill Expert Survey (CHES), in which experts place the parties of Europe, works the same way.

Three things are our own choice: the selection of the seven axes, the five steps and the measure. We measure how far a point wants to change today's state. There is no fixed model for that.

You can read the complete instructions to the AI word for word: The prompt.

The cross-check against an expert survey

We compared our values with the CHES of 2024. It covers eight of the parties that have a program for the 2025 federal election: AfD, BSW, CDU and CSU, FDP, Freie Wähler, Grüne, Linke and SPD.

The CHES measures where a party stands. We measure how far a program wants to move from today's state. So the size of the values cannot be compared, but the order of the parties can. Eight parties make 28 pairs. The table shows for how many pairs our order is the same as the experts'. Next to it is the correlation: 1 means complete agreement, 0 means none.

Axis	Same order	Correlation
Migration	27 of 28 pairs	0.96
Economy	25 of 28	0.95
Society	23 of 28	0.92
Europe	23 of 28	0.85
Foreign policy	24 of 28	0.82
Climate	22 of 28	0.82
Civil liberties and security	16 of 28	0.33

As of 4 October 2026.

Six of the seven axes agree well with the experts. "Civil liberties and security" agrees poorly: 16 of 28 pairs is hardly more than chance would give. The reason lies in the procedure. A program lists single plans, and here every plan counts the same. Parties that want to govern usually promise many posts for the police and the courts. Experts, by contrast, rate a party's basic stance. So read this axis as the balance of the plans in the program, not as a basic stance. For the same reason the compass shows the society axis vertically.

Three limits apply to the whole comparison:

• Eight parties are few. The numbers show a direction and are no proof.
• Only the programs for the 2025 federal election are compared, and the survey is from 2024. The state programs are not part of the comparison, because the CHES places national parties.
• On foreign policy the CHES asks only about the relation to Russia, and few experts answered that.

We have not yet measured how much a placement varies when the AI reads the same program a second time.

Sources: Jan Rovny and others, "The 2024 Chapel Hill Expert Survey on political party positioning in Europe", Electoral Studies 97, 2025. Gaël Le Mens and Aina Gallego, "Positioning Political Texts with Large Language Models by Asking and Averaging", Political Analysis, 2025. The Manifesto Project: manifesto-project.wzb.eu.

Why a placement can be surprising

We place only what the election program says. The picture many people have of a party also comes from other things, such as speeches, its work in government and its history. So our placement can differ from that picture without being wrong. Common reasons:

• A program shows what a party promises. It stresses some topics and leaves out others.
• There is no single "left" or "right" for everything on this site. In everyday speech, "right" often means migration or society, and "left" often the economy. Here each of these topics has its own axis, and a program can stand somewhere else on each.
• A party's name and history do not necessarily describe its current program.
• On "Civil liberties and security", every single plan counts. A program with many points on police and criminal law therefore sits closer to "security and order first", even if the party is seen as liberal. Points against surveillance, in turn, move a program towards "civil liberties first".
• The placement measures how strongly a program wants to change the current state. A party that defends what exists therefore sits closer to the middle, even if it holds firmly to it.

Every point comes with the quote and a reason. So you can check every placement yourself.

Placements are assessments

A placement is our assessment of a point by the method this page describes. It is a value judgment, not a statement of fact. The values of a program or topic are averages of such assessments. A placement says nothing about whether a demand or a party is compatible with the German Basic Law (Grundgesetz).

Reporting errors

On the topic pages, every point has a "Report an error" button below it. It sends a message about exactly that point, for example when a summary says something wrong or a placement does not fit. For anything else, use the contact form.

Sample data

Not every program has been analysed yet. For the others, the site shows made-up sample programs. They only serve to test the structure and presentation and do not reflect any party's positions. Pages with sample data carry the note "Sample" at the top and are blocked for search engines.