Google AI search summaries vary across political and news searches, study finds
Author ADM+S Centre
Date 21 May 2026
Google’s AI-generated search summaries are changing how users encounter political and news information online, with new Australian research finding the results are neither uniform nor neutral.
A new study has produced the first systematic mapping of Google’s AI Overviews (AIOs), the automatically generated summaries that appear at the top of search results, replacing traditional ranked links with a single AI-generated response.
AI Overviews vary by political context
The research conducted by PhD Student Shir Weinbrand from the ARC Centre of Excellence for Automated Decision-Making and Society at QUT, found nearly half of all political and news-related Google searches examined generated an AI Overview, despite public suggestions from Google that the feature would be limited for hard news queries.
“We are at a pivotal moment in the transformation of search, as AI Overviews reshape how political and news-related information is surfaced online,” said Shir.
“While Google states that AI Overviews are not shown on news queries, our findings suggest a more complex deployment in practice. This highlights the need for further research into their implications for information visibility, platform transparency, and public knowledge.”
The research analysed more than 17,000 Google search results across desktop and mobile devices in Australia between April and June 2025, covering searches related to political issues, elections, politicians and political news.
It found that political issue queries triggered more comprehensive AI-generated summaries with richer structures and more extensive explanations, potentially reflecting design priorities to scaffold complex topics. However, these explanations often appeared without prominent inline source links, raising questions about transparency and traceability in political information provision.
In contrast, news on politics and political process queries often triggered leaner formats, aligning with concerns that news and political information exposure can vary systematically with interface design and query context.
Concerns over “algorithmic gatekeeping”
The study warns these variations might amount to a new form of algorithmic gatekeeping.
This gatekeeping extends beyond deciding which sources are surfaced to shaping how their content is synthesised, packaged, and displayed.
Previous research has shown that search engines can affect exposure to diverse news content, influence attitudes on issues, and shape political learning.
With the addition of AIOs, these effects may be amplified, as users are increasingly presented with a pre-framed response rather than a list of links to explore. Investigating the components and structure of AIOs is therefore crucial.
“AI Overviews are neither uniform nor neutral,” the study found.
“Their structure varies systematically according to query characteristics.”
The study suggests this may reflect an intentional design decision by Google to treat ongoing news events more cautiously while providing richer summaries for longer-running political topics.
Some political searches return no AI summary
The study also identified cases where Google declined to generate AI summaries altogether, instead displaying the message: “An AI Overview is not available for this search.”
These refusals appeared most often for searches involving political figures and sensitive political topics.
The study argues the wording lacks transparency because users are not told whether:
- insufficient reliable information exists,
- technical limitations occurred,
- or Google deliberately chose not to generate a summary.
Implications for public knowledge
The findings raise broader concerns for journalism and democracy as AI-generated search summaries become more common.
Recent research cited in the study suggests users are significantly less likely to click traditional news links when AI summaries appear prominently at the top of search results.
This could reduce traffic to news publishers while concentrating even more informational power within major technology platforms.
The study also warns AI-generated summaries may obscure the origins of information by blending material from multiple sources into a single narrative, making it harder for users to identify where claims come from.
New taxonomy for auditing AI search
The study also developed the first detailed taxonomy of AI Overview interface elements, providing a framework for future auditing and cross-platform comparisons of AI-powered search systems. An interactive dashboard is publicly available for exploring the data.
The dashboard brings the data to life through interactive features spanning AIO status over time and across queries, interface anatomy and live taxonomy, and AIO structural types.
The study provides an important framework for cross-platform auditing of AI-powered search systems, equipping researchers in HCI and political communication with tools to better understand how generative search shapes information exposure – and calls for greater transparency around interface design, algorithmic accountability, and potential biases in how political information is presented..
Design recommendations
This study also highlights several design elements that are currently absent from Google’s AIO interface. For instance, expandable sections toggles could allow users to collapse or expand segments, facilitating selective reading and reducing cognitive load.
Displaying the number of sources associated with each citation icon could enhance transparency and media pluralism, while ranking cited sources by authority could strengthen accountability and fairness in AI-mediated information environments. Incorporating such elements would improve user control, transparency, and trust in AI-generated political and news information.
The study concludes that the shift from traditional search to AI Overviews will shape public knowledge and civic life for years to come, demanding sustained attention from researchers, policymakers, and society at large.
This research has been undertaken as part of the ADM+S Australian Search Experience 2.0 project.
Visit the AIO interactive dashboard
Read the research paper A search changer: auditing Google’s AI overviews interface in political and news search
