Why this toolkit

Extreme weather events are fuelling dangerous misinformation and disinformation in communities across the United States. These crises can be weaponised by domestic or transnational actors, leading to chaos that endangers lives and hinders local responses. This toolkit positions information resilience as a key element of disaster preparedness, response, and recovery. It is designed primarily for emergency managers, resilience officers, and public information officers. Between 2022 and 2024, NOAA recorded 73 distinct events that cost an estimated $461 billion to the economy. The annual number of billion-dollar disasters increased from nine in 1980 to an average of twenty-six in the last three years.

Understanding the threat

Information threats around extreme weather generally fall into three categories: environmental (attacking science); (geo)political (attacking responses); and hate, extremism, and conspiracies (attacking people). Environmental threats divorce weather from its drivers or exploit scepticism about climate change. (Geo)political threats attack preparedness by linking disasters to unrelated policy issues, such as government spending or migration. The third category scapegoats minorities or propagates theories about elite control. The rapid growth of Large Language Models (LLMs) heightens these challenges, as AI tools help to create and disseminate false content at scale. Specific harms include ‘hallucinations’, where AI generates incorrect information, and ‘LLM grooming’, where models are poisoned upstream to make certain narratives more visible.

Triaging information threats

A risk matrix is essential for triaging threats in real-time. Harm is assessed based on the immediacy, severity, and prevalence of impacts, weighed against available resources. The ‘Ocean’ model categorises misinformation by speed: ‘Waves’ (viral surges lasting hours to days), ‘Tides’ (event-driven surges lasting weeks), ‘Currents’ (enduring narratives flowing for months or years), and ‘Gulf Streams’ (continuous transnational operations). High-immediacy and high-severity risks, such as threats to life or damage to critical infrastructure, require downstream interventions like official debunking, paid media, and escalation to law enforcement. Lower-level risks are managed through upstream interventions like staff training and staff skill development.

Assessing readiness

Information resilience starts with a stock-take to evaluate expertise, protocols, and political backing. Organisations should map tangible assets, such as funding and equipment, alongside intangible assets like relationships and trust. The report recommends assembling a working group across government, private, and civil society sectors to run red-team exercises. These exercises use plausible scenarios, such as intensified wildfire seasons or storms during major cultural events, to plot best- and worst-case trajectories. This helps identify chokepoints and develop contingency plans that participants are empowered to operationalise before a crisis occurs.

Partnerships for networked information resilience

Information resilience is built on networks, not newsfeeds, as data shows Americans trust personal connections most. More than 4 in 5 Americans view personal connections as a trustworthy source, while only 15% have high trust in online influencers. Trust in government and media remains low, around 40%. The toolkit suggests investing in ‘truth multipliers’—trusted community figures such as doctors, teachers, faith leaders, and even hairdressers or sports coaches. These partners can act as ‘canaries in the coal mine’ to provide early warnings. Effective partnerships require clear incentives, including micro-grants, professional development, or visibility in the community. Partner safeguarding is also critical to protect these individuals from online and offline backlash.

Defences against information manipulation

Pre-bunking, or inoculation, acts as the first line of defence by building resistance before people encounter false claims. The ‘truth sandwich’ model is recommended: leading with facts, warning of the myth, explaining the fallacy, and restating the facts. To counter the ‘zero click’ world created by AI, the report suggests Generative Engine Optimisation (GEO). This involves making high-trust content more visible to AI chatbots by hosting Subreddits, updating Wikipedia pages, and using ‘AI-readable’ structured content like FAQ schemas. Governments are also encouraged to bypass newsfeeds by curating hyperlocal direct channels, such as WhatsApp groups, Nextdoor forums, and text alerts.

Concluding thoughts

Building information resilience is a process that must feature in wider roadmaps. While digital platforms are important, they are an unequal playing field where low-trust information often has a 3 to 35 times engagement advantage over factual posts. Local governments must avoid letting a vocal fringe confer legitimacy and should instead focus on the ‘silent majority’. Both information and climate resilience must centre agency as a key theme, leaving communities feeling united by a common struggle and empowered to act. Rebuilding the foundations of epistemic integrity is necessary both upstream, by reshaping discourse environments, and downstream, by empowering individuals to resist and correct disordered information.