Generative AI gives grassroots advocacy groups new tools for civic action

A new study published in AI & Society reveals that generative artificial intelligence (genAI) could help grassroots community groups scrutinize complex planning documents, expand public outreach and strengthen advocacy campaigns, whilst creating serious concerns over accuracy, privacy, misinformation, intellectual property and trust.

The study, titled Digital civics in action: the promise and emerging practices of generative AI in grassroots community advocacy, examines how tools such as large language models and image generators can support civic action by local community groups, drawing on interviews and a co-design workshop conducted with Brisbane Residents United and affiliated grassroots advocacy organizations in Australia.

Grassroots groups face civic barriers that GenAI may help reduce

The authors argue that formal engagement channels often leave residents feeling unheard or excluded, with public consultation processes constrained by institutional power, tokenistic participation, limited civic education and difficult access to decision-makers.

Grassroots advocacy groups have stepped into that gap. They monitor planning decisions, challenge public authorities, organize residents, write submissions, run campaigns and try to translate technical planning issues into language that communities can understand. However, these groups often work with limited time, limited funding and heavy dependence on volunteers. Many members are older, and some groups reported weak social media capacity, uneven digital literacy and difficulties reaching younger residents.

The study identifies three linked areas of civic practice where community groups operate: scrutiny, outreach and advocacy. Scrutiny involves reading, questioning and interpreting information from councils, developers and governments. Outreach involves informing residents, expanding awareness and mobilizing support. Advocacy involves turning that concern into petitions, submissions, campaigns, public statements and pressure on decision-makers.

The authors found that community groups face barriers across all three areas. Residents may lack time, fear consequences for speaking out or stay at surface-level engagement. Advocacy groups may struggle to divide tasks among volunteers, maintain momentum or keep up with procedural requirements. Accessing public information can also be difficult. Participants described government systems as hard to navigate, with planning documents often long, technical and scattered across multiple files.

According to the study, GenAI is a possible support tool for grassroots action, especially when used by communities to strengthen existing practices. The key question is how these tools can be adopted strategically without worsening existing power imbalances.

The research used a participatory, design-led method. The team worked with Brisbane Residents United, a peak body representing local citizen action groups in Brisbane. Semi-structured interviews were conducted with representatives from five grassroots groups, while a half-day co-design workshop included nine participants from different community organizations. The workshop first mapped existing advocacy practices and then explored possible GenAI use cases across scrutiny, outreach and advocacy.

The approach allowed the authors to ground the findings in actual community practice rather than abstract technology promises. The participants were not treated only as interview subjects. They helped identify use cases, risks and practical conditions for GenAI adoption in grassroots advocacy.

AI can simplify planning documents and strengthen campaign communication

Grassroots groups often have to review large planning documents, development applications, policy files and technical reports within short public consultation periods. These materials may run across many files, include hundreds of pages and use professional language that is difficult for residents to interpret quickly.

Participants said GenAI could help summarize complex documents, extract key issues, translate planning jargon and support quick question-answer interactions with dense material. This could give volunteer groups a faster first pass through documents that previously required days or weeks of manual review. The tool could also help identify policy implications, compare proposals and assist groups in preparing responses before consultation deadlines close.

The study also identifies GenAI support for comparative policy learning. Community groups may want to understand how similar planning issues have been handled in other cities or countries. GenAI tools can help locate examples, compare approaches and generate summaries that support local learning. This could be especially useful when groups need to develop informed arguments against a proposal or show that other policy paths are possible.

In outreach, participants saw GenAI as a tool for content creation and public communication. Grassroots groups already rely on Facebook, newsletters, email, websites, community meetings and local events to build awareness. But producing clear, compelling and consistent content takes time. GenAI could help draft social media posts, write emails, prepare meeting notes, generate campaign captions and support media releases.

Image generation also emerged as a potential outreach tool. Participants saw value in using GenAI to create visuals for social media, websites, blogs, posters and campaign material. These visuals could help explain complex planning issues more quickly and draw public attention to developments that may otherwise remain buried in technical documents.

A related use case is alternative scenario generation. Grassroots groups are often framed as oppositional or obstructionist when they challenge planning proposals. Participants saw GenAI as a way to help communities present constructive alternatives, including images and narratives of more sustainable, inclusive or liveable urban futures. This could help groups move beyond saying no to a project and instead show what a better outcome might look like.

In advocacy, participants identified GenAI’s potential to support petitions, public submissions, campaign strategies and formal statements. These tasks often require polished language, clear argumentation and policy relevance. GenAI could help draft material that community groups then review, verify and adapt. The authors suggest this could help volunteer-based organizations save time while improving the clarity and reach of their interventions.

The study also points to a broader democratic issue. Developers, consultants and public agencies often have paid staff and professional communications support. Grassroots groups rarely have similar resources. GenAI may help reduce some of that imbalance by giving community groups access to faster document processing, content generation and visual communication tools. But the study is careful not to present this as a full solution to structural inequality.

Participants stressed that GenAI still requires skilled human use. Effective prompting, domain knowledge, fact-checking and critical review remain necessary. The tools can assist scrutiny, outreach and advocacy, but they cannot replace local knowledge, legal understanding, planning expertise or community judgment.

Risks around privacy, misinformation and ownership demand safeguards

GenAI could create new risks for the same communities it aims to support, the study clearly warns. Participants raised concerns about confidentiality, especially when grassroots groups deal with sensitive information, internal strategy or personal details. Entering such information into AI tools may create uncertainty over storage, access and future use.

Accuracy was another major concern. GenAI systems can produce incorrect information or present invented details with confidence. In advocacy, this risk is serious because inaccurate claims can damage credibility, confuse residents and weaken public submissions. The authors stress the need for human oversight and verification, especially when AI-generated material is used in planning disputes or formal civic processes.

Misinformation and manipulation were also among the key concerns. Participants worried that GenAI could distort issues, exaggerate claims or be misused to spread misleading campaign material. This risk applies not only to grassroots groups but also to more powerful actors. The study notes that GenAI can be a double-edged tool: it may help communities understand complex documents, but it could also enable governments, developers or consultants to produce even larger and more complex materials that overwhelm lay readers.

Intellectual property concerns were also raised. Community groups may upload their own campaign content, strategies, documents or images into AI platforms without full clarity over how those inputs might be reused or adapted. Participants expressed concern over ownership and control of content generated or shared through GenAI systems.

GenAI use in grassroots advocacy should be guided by transparency, accountability, privacy safeguards and ethical rules, the authors insist. This includes clearly identifying when AI has been used, assigning responsibility for checking outputs, avoiding sensitive data uploads where risks are unclear and building community capacity through digital and media literacy training.

The study also calls for collaborative, multi-actor approaches to AI governance. Community groups may need support from AI experts, legal advisers, planners, designers and civic organizations to build safer workflows. Human oversight is necessary, but the authors warn that oversight cannot be symbolic or reactive. It must be built into the full lifecycle of AI use, from selecting tools and entering data to checking outputs and publishing campaign material.

It's important to mention that the research was based on a small, context-specific sample of Brisbane advocacy groups, and the co-designed use cases were not implemented and tested in real campaigns. The study does not empirically measure whether GenAI improves advocacy outcomes or whether proposed safeguards work in practice. Future research, the authors say, should prototype and evaluate these use cases across different community, political and organizational contexts.