Anyone looking to Google's AI Overviews for information during upcoming state elections should remain critical. That's what an investigation by Algorithmwatch reveals, having systematically analyzed in July 2026 how Google handles AI-generated answers to election-related queries.
Key Facts
- Algorithmwatch analyzed nearly 4,500 search queries about the 2026 state elections in Saxony-Anhalt, Mecklenburg-Vorpommern, and the Berlin parliamentary election
- Google displayed an AI Overview for only 39.1% of election-related queries – compared to 65.3% for non-political comparison queries
- The analysis suggests Google makes deliberate decisions about when to show AI answers, which sources to cite, and how to formulate content
- Researchers used Google's "Search Researcher Result" API interface for the study
Significantly Fewer AI Answers on Elections
Algorithmwatch gained access to Google's internal software interface and systematically posed questions about the three election campaigns. In parallel, researchers created a control dataset with non-political search queries – such as travel destinations and activities in specific regions.
The result is striking: For queries about the 2026 state elections, Google displayed an AI Overview in only 39.1% of cases. For non-political queries, the rate was significantly higher at 65.3%. This raises the question of whether Google is deliberately exercising restraint here – or whether quality assurance for politically sensitive content is working too aggressively.
Source Selection and Wording Under Scrutiny
The analysis doesn't stop at the frequency of AI answers. Algorithmwatch also examined which sources Google cites and how content is formulated. The findings suggest Google is not proceeding neutrally:
| Aspect | Finding |
|---|---|
| Display Rate | Election topics: 39.1% vs. non-political: 65.3% |
| Source Selection | Differences in selection and weighting |
| Formulation | Content sometimes appears non-neutral |
This means: Google apparently makes deliberate decisions about whether to display an AI Overview at all, which sources to cite, and how to formulate content. This is particularly problematic for elections – where information access and neutrality are central.
What This Means for Users and Regulators
The findings raise questions about transparency and control. Users relying on AI Overviews for voting decisions could unknowingly encounter selection mechanisms they don't understand. At the same time, it becomes clear: Google's algorithms make decisions on politically sensitive content that are not simply technically neutral.
For German companies and platforms, this is an important lesson. Regulation will increasingly impose transparency requirements for AI systems handling political and socially relevant content. Anyone operating such systems will soon need to demonstrate that source selection, display rates, and formulations are not discriminatory or manipulative – especially during elections and political processes.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
