Search Climate Insights Directory

The report finds that text-based GenAI misinformation currently poses a higher persuasive risk than visual misinformation. It identifies preemptive debunking as the most consistent intervention for reducing perceived credibility, while content labeling is effective only when standardized. The analysis highlights a critical evidence gap, as existing research is heavily concentrated in English-speaking and high-income countries.

The report finds that textual GenAI misinformation is increasingly perceived as credible over time, while visual GenAI misinformation, such as deepfakes, shows a decline in perceived credibility. Preventive corrective information and content labeling are identified as effective countermeasures, though their impact varies by context.

The report evaluates ten distinct indexes designed to diagnose the resilience or vulnerability of political information environments to mis- and disinformation. It develops a typology of these tools—categorizing them as advertising-oriented, policy-driven, academically oriented, or public outreach—and analyzes their conceptual foundations, data collection methods, and scoring architectures. The review finds that most indexes rely on proxy indicators (such as institutional trust or media freedom) rather than direct measurement of false content. Comparative case studies reveal that while some indexes converge on structural constants in advanced democracies, others diverge sharply due to different operational definitions of 'resilience,' such as distinguishing between reactive mobilization and long-term structural capacity. The document concludes that current measurement is fragmented and recommends a more unified architecture that integrates structural, behavioral, and platform-specific dynamics.

The document serves as a technical appendix to a meta-analysis on generative AI misinformation. It details the risk of bias assessment for various randomized controlled trials using the Cochrane RoB 2 protocol, finding that most studies scored 'low' to 'moderate' for bias. It also describes tests for publication selection bias—including Egger's test and the Duval and Tweedie trim-and-fill method—across samples focusing on visual misinformation, corrective information, and content labeling, concluding that while some small-study effects may exist, they do not fully explain the observed effects. Additionally, it provides supplementary figures on the perception of GenAI misinformation and the efficacy of corrective information.

The report finds that text-based generative AI misinformation currently poses a greater persuasive risk than visual misinformation. It identifies preventive corrective information as the most consistently effective intervention to reduce the perceived credibility of such content, while noting that content labeling is promising but highly dependent on consistent design and implementation. The authors highlight a critical evidence gap, as current research is heavily concentrated in English-language and high-income countries.

The document analyzes the impact of generative AI misinformation, finding that textual misinformation currently poses a higher persuasion risk than visual misinformation. It highlights a significant global data gap, as research is heavily concentrated in high-income, English-speaking countries. Among countermeasures, preventive corrective information is identified as the most consistent intervention for reducing perceived credibility, while content labeling is effective only when applied consistently and with rigorous design.

The report finds that generative AI-produced textual misinformation currently poses a higher persuasion risk than visual misinformation. It identifies preventive corrective information as the most consistently effective intervention for reducing the perceived credibility of such content, while content labeling shows promising but inconsistent results depending on design. The analysis highlights a significant global evidence gap, as most research is concentrated in high-income, English-speaking countries.

The report analyzes 60 effect estimates from 24 peer-reviewed studies (33,801 participants) published between 2018 and 2025 to evaluate generative AI (GenAI) misinformation. It finds that GenAI-generated text currently poses a higher risk of persuasive impact than visual misinformation. The most consistently effective countermeasure is the preemptive provision of corrective information, while content labeling shows variable effectiveness depending on design and application. The authors highlight a significant global knowledge gap, as research is heavily concentrated in English-speaking, high-income countries.

The report finds that text-based generative AI misinformation poses a higher persuasive risk than visual misinformation. It identifies a significant knowledge gap due to research focusing primarily on high-income, English-speaking countries. Among countermeasures, preemptive corrective information is the most consistent and effective intervention, while content labeling shows promising but inconsistent results depending on design and application.

The document examines how AI can support peacebuilding through early warning systems, digital dialogue, and humanitarian mapping, while warning against risks such as digital surveillance, misinformation, and algorithmic bias. It argues that AI must be treated as a supplementary tool governed by human rights and 'do no harm' principles, emphasizing the need for localized models and cross-sector collaboration to prevent the technology from exacerbating instability in fragile regions.

The document identifies a strategic shift from outright climate denial to 'strategic doubt,' where influential economic and political actors—including fossil fuel companies and right-wing populist movements—deliberately spread misleading information to obstruct climate action. It highlights that policymakers are primary targets of these campaigns, which are disseminated through traditional media, social media, and non-public corporate channels. The report notes a significant research gap in the Global South and proposes four policy interventions: mandatory standardized carbon reporting, litigation against greenwashing, the creation of cross-sectoral voluntary alliances, and the enhancement of scientific and media literacy for the public and officials.

The report analyzes 215 cases of GenAI use in 2024 elections across 50 countries, finding that 80% of these countries experienced GenAI-related incidents. The majority of these cases (90%) involved content creation, such as deepfake audio and video, and 69% were classified as harmful to the electoral process. A significant portion of the content (46%) had untraceable origins, while 20% was attributed to foreign actors, all of whom were found to be acting with harmful intent. The document advocates for increased transparency, disclosure mandates for AI-generated political content, and improved detection mechanisms to protect democratic integrity.

The report finds that 80% of countries with competitive national elections in 2024 experienced GenAI-related events, with 69% of these events having a harmful impact. The majority of GenAI use focused on content production, and nearly half of the events had unknown sources.

The 2025 Expert Survey on the Global Information Environment, involving 438 researchers across 76 countries, examines the state of 'epistemic security'—the reliability of information supply chains. The report finds that while social media can increase political participation, it also drives polarization and distrust in institutions. A significant 66% of surveyed experts reported experiencing harassment, intervention, or chilling effects in their scientific work, with funding constraints and data access remaining primary barriers. Regarding AI, there is strong expert consensus (60%) that AI-generated content should be clearly labeled, and 83% believe such labeling would improve the information environment. The document advocates for treating the information environment as critical infrastructure, recommending independent algorithmic auditing and stronger protections for researchers and journalists.

The document examines the dual-use nature of AI in peacebuilding, highlighting its ability to forecast violence and scale civic inclusion while warning against risks like digital surveillance, algorithmic bias, and the erosion of human-led diplomacy. It argues for a hybrid approach that combines AI tools with local expertise and strict ethical governance to ensure technology supports stability rather than deepening conflict.

The report analyzes 215 recorded GenAI incidents across 50 countries during 2024 elections, finding that 80% of these countries experienced such incidents. GenAI was primarily used for synthetic content production, with 69% of cases deemed harmful to electoral integrity. Attribution is a significant challenge, as 46% of sources remained unknown, while political candidates and foreign actors accounted for 25% and 20% of incidents, respectively.

The report identifies a crisis of information integrity about climate science, characterized by a shift from outright denial to more nuanced skepticism and 'response skepticism' regarding climate solutions. It documents how powerful economic interests, particularly the fossil fuel industry, and political actors use strategies like greenwashing, delay, and obstruction to undermine scientific consensus. The review highlights the role of social media algorithms, bots, and 'source media' (such as corporate reports) in amplifying misleading narratives. It concludes that while individual-level interventions like inoculation and education are valuable, systemic governance through public regulation, standardized corporate reporting, and litigation is essential to restore information integrity.

The International Panel on the Information Environment (IPIE) argues that the UK's Data (Use and Access) Bill must move from discretionary to obligatory language regarding researchers' access to platform data. The IPIE recommends that the Secretary of State be required to set regulations for data access, that platforms bear the costs of this access to ensure equity among researchers, and that enforcement mechanisms be strengthened to prevent platforms from using privacy concerns as a pretext to deny data. These measures are presented as essential for fulfilling the Online Safety Act's goal of making internet services safer for individuals in the United Kingdom.

The IPIE analyzed GenAI usage across 50 countries with competitive national elections in 2024. Findings show that GenAI was reported in 80% of these countries, primarily for content creation (90% of cases). While 69% of instances served malicious ends, 16% were benevolent. The study highlights a significant challenge in actor identification, with 46% of cases remaining unclear, and notes that foreign actors were involved in 20% of reported incidents, all for nefarious purposes.

The report presents findings from the 2025 IPIE Expert Survey, highlighting a growing pessimism about the future of the information environment. It identifies the availability of accurate information and diverse voices as essential for a healthy environment, while citing lack of platform accountability as a primary threat. The document also examines the mixed impact of social media, the negative perceptions of generative AI, and the increasing barriers faced by researchers, particularly regarding funding and academic freedom.