Can AI strengthen democracy? A data collection on AI projects aiming to serve democratic processes

Artificial intelligence is often presented as a tool that could potentially strengthen democratic processes, but so far there is little empirical evidence either for or against these claims.  Organizations proposing such AI tools are the first focus of the Impact AI project, a research project that is assessing the real-world impact of AI projects oriented towards public interest and sustainability. However, there are insufficient data collections that allow an overview of this new application field of AI. Thus, this article introduces a dataset displaying 98 AI projects that explicitly claim to serve democratic purposes. The dataset is intended as a resource for researchers seeking to empirically investigate these claims and assess their societal implications. The article outlines how the dataset was developed, what information it contains and how it can contribute to building evidence and transparency concerning the use of AI for democracy.  

Around the globe, a whole generation of young people is already growing up using a multitude of AI-driven tools. Now specific applications that aim to strengthen democratic processes are emerging. In a time in which democracy is in crisis, many initiatives focus their efforts on restoring and supporting democratic practices or on analysing the democratic crisis and discussing paths for stabilization. The AI Action Summit in 2025 hosted sessions around the idea that AI systems could support this struggle. Their claim was that in the future  AI applications could become relevant for participation and democratic practices and governance. 

Serving democracy with AI?

One recent example from Columbia is the AI avatar “Gaitana” (Lozada & Chandran, 4 March 2026). Created by members of the Zenú Indigenous community, the bot was designed to digitally represent two human candidates running for parliamentary seats in the March 2026 elections. If elected, the two candidates would have deferred all legislative decision-making to the platform, which uses AI to gather opinions and build consensus among community members, with blockchain and smart contracts to verify the process. But neither candidate won a seat, so that arrangement was never put into practice (Acosta, 17 March 2026).

 Besides this example of augmenting  political representation, there are a number of other applications of AI in democratic processes for instance with a focus on investigating hate speech (Landesanstalt für Medien NRW, 7 February 2025), facilitating online deliberation, monitoring public spending or supporting negotiations as in the case of an application which was tested during the Cop30 Climate Conference in Brazil. This new application field of AI in democratic processes raises many questions. The central one is, if the claims hold water and how such applications of AI might be impactful and supportive of democratic deliberation and governance.

Assessing the impacts

In our transdisciplinary research project Impact AI at the Alexander von Humboldt Institute for Internet and Society (HIIG), we investigate the impact of AI projects intended to serve society and the planet. Together with our civil society partners, Greenpeace and EconGood, we have decided to make ‘AI for democracy’ our first focus area. This means that we are assessing three AI applications in this field. We chose this topic not only because it so prominently features in global political forums but also because insights on how AI applications affect democratic practices are crucial for the future of our society. In our research, we are committed to understanding the impacts of such projects. We especially emphasise how they perform in regards to sustainability and aspects representing public interests. Our research team developed a holistic impact assessment method to answer these questions and provide more transparency in this timely discourse. 

Closing in on the field

To select cases in the field of AI for democracy for our impact assessment, we first mapped the field and collected as many projects as possible from which to choose from. The “AI for Democracy Dataset” we publish here is the result of this mapping exercise. We make our collection of use cases in the field of AI for democracy publicly available as we believe it will be of interest to civil society and other researchers, providing a basis to ask questions on ‘AI for democracy’. For the projects in the dataset we cannot draw the assumption that these are actually supporting democracies. We only acknowledge that they make a claim towards this goal, which is an important difference. We cannot affirm or deny any positive or negative impacts of the projects in this data collection. Rather, the projects in this data collection self-describe as having a positive impact on democracy. Also, even though the dataset is quite extensive, it is by no means exhaustive or representative of the field.

How was the data collected?

We gathered the ‘AI for Democracy Dataset’ from multiple sources. The herein published dataset builds on previously collected data by the Public Interest AI research group published on Hugging Face. Additionally, it also includes data from two European initiatives which emerged around the AI Action Summit in Paris 2025. First, a number of projects presented by the AI for Democracy – Democratic Commons initiative and second, some of which answered a call by the Paris Peace Forum  for AI for public good, of which 50 projects were selected from 770 submitted overall. We took these 50 AI projects as a starting point, identified 24 of them as having a connection to democracy, and added them to our dataset.

Beyond these sources, we methodologically assembled other relevant AI projects through a purposeful sampling approach (Bouncken et al., 2026) building on a web search, where we examined the results of specific search queries (e.g. ‘AI for democracy’) and a network-based search method utilizing professional networks and contacts to identify other relevant AI projects.

What’s in the dataset?

Any project identified through these interactions was documented and vetted against the minimal definition in the dataset, i.e. making a claim to use AI for democracy based on its online available documentation. The resulting ‘AI for Democracy Dataset’ includes 98 AI projects with an explicit claim to serve democracy. We investigated every single project’s website before including it in the dataset. In addition to every project’s basic information and according to our research efforts, we decided on additional characteristics for documentation (please find the full description in the dataset’s readme).

We started with each project’s name. Next, we list the SDG they relate to. Though all projects have a connection to democracy, they can still have specific focal points and therefore relate to different SDGs. While some AI projects gathered in this dataset clearly name the SDGs as point of reference, we categorised others in this regard with one or two SDGs which we considered best matching the project goals. We also attributed hashtags aimed at narrowing down the respective application’s field of engagement, thus providing more details. Also, we added information on whether a project operates for-or-nonprofit, or as a research project. Adding information on the AI/Technology used, we share our best-guess, building on the information provided by each project, before naming their owners, and the project’s activity status. To every project’s location, we added the application’s targeted region, if available. If we were able to identify sources of funding, we included them in the dataset as well. Besides this systematisation of information, the dataset provides a weblink to each of the project’s websites, thus making it accessible for more thorough investigation. Each project’s description, and all other information gathered in the dataset originate from the corresponding website. It is important to note that the information given might have changed since then and refers solely to the time of each project’s inclusion into the dataset.

Food for future research

At this point we cannot provide a deep analysis into all of the dataset’s projects, but it may help other researchers understand the landscape of ‘AI for Democracy’ and identify questions for further research. 

Based on the AI project’s descriptions and websites alone, it is hard to determine which projects are really active (still or yet) and what they have achieved. A follow-up study revisiting these projects in a year’s time could shed light on the noticeable tendency of AI projects to wind down once funding ends.

Looking at our dataset, we observed that a substantial number of projects rely on Natural Language Processing for their purposes. This likely reflects the discursive nature of much of this work: many projects engage with democracy through language — summarising group discussions, detecting misinformation, or supporting mediation. Future research could investigate this further and examine how the specific uses of AI manifest a specific understanding of democratic processes and how they position the role of AI within it. 

Only a beginning

AI for democracy is a rather young and still-evolving application field. Many of the projects we encountered are still testing their approaches, iterating on their methods, and changing strategies to create impact. With this dataset, we hope to invite other scholars to join in building an evidence base, sharing learnings, and fostering transparency around the use of AI for democratic processes.

We look forward to reporting on individual applications throughout 2026 and are curious to see this field develop. In a time of rising uncertainty and global competition, open and respectful exchange and critical reflection, as well as international cooperation are the conditions under which technology, and AI in particular, can be developed for the benefit of all.

References

Acosta, A. (2026, March 17). “Gaitana IA”: The AI candidate that ran in Colombia’s elections. Latin America Reports. https://latinamericareports.com/gaitana-ai-the-ai-candidate-that-ran-in-colombias-elections/13929/

AI for Democracy – Democratic Commons. (2024). AI for Democracy: Democratic Commons. https://about.make.org/democratic-commons/landing-page

Bouncken, R. B., Czakon, W., & Schmitt, F. (2026). Purposeful sampling and saturation in qualitative research methodologies: Recommendations and review. Review of Management Science, 20, 579–615. https://doi.org/10.1007/s11846-025-00881-2

Élysée. (2025). Artificial intelligence action summit. https://www.elysee.fr/en/emmanuel-macron/artificial-intelligence-action-summit

Landesanstalt für Medien NRW. (2025, February 7). Für mehr Sicherheit im Netz. https://www.medienanstalt-nrw.de/presse/pressemitteilungen/pressemitteilungen-2025-02/fuer-mehr-sicherheit-im-netz.html

Lozada, M., & Chandran, R. (2026, March 4). An AI avatar is running to represent Indigenous voters in Colombia. Rest of World. https://restofworld.org/2026/ai-avatar-colombia-political-candidate/

NegotiateCOP. (n.d.). NegotiateCOP. https://negotiatecop.org/

Paris Peace Forum. (2024). Call for AI projects. https://parispeaceforum.org/call-for-ai-projects/

Serenata de Amor. (n.d.). Serenata de Amor: Artificial intelligence for social control of public spending. https://serenata.ai/en/

Stanford Online Deliberation. (n.d.). Online deliberation platform. https://stanforddeliberate.org/