AI crisis communication: what it is and how it works

What AI crisis communication actually means

By AI crisis communication we mean the use of artificial intelligence within the communication around a crisis: from a major fire to a cyber incident, from flooding to a data breach. It is not a new discipline replacing the old one, but a layer of tools that operates at three points in the process.

First, there is AI that speeds up the picture of the situation: monitoring social media, news sites and incoming reports, so a crisis team sees faster what is happening and how a situation is developing. Second, there is AI that helps write the messages: drafting a first version of a press statement, a social post or an internal update, which a communication adviser then checks and sharpens. Third, there is AI that speaks to the public and other stakeholders directly, for instance through a chatbot answering frequently asked questions during a crisis, so phone lines and service desks are freed up for the questions that genuinely need a person.

These three run into each other. A municipality can use AI to monitor social media for signals about a major power outage, have a draft message prepared for the moment the facts are confirmed, and point residents to the latest update through a chatbot. The pattern is always the same: AI accelerates and supports, people judge and decide.

Established methodology stays in charge

Anyone introducing AI into crisis communication is well advised to get the existing methodology straight first, because AI changes very little about it. In the Netherlands, crisis teams generally work to the BOB decision model: Beeldvorming (building the picture: what do we know, what is fact and what is assumption), Oordeelsvorming (judgement: what does this mean for our objectives and dilemmas) and Besluitvorming (decision: which action do we choose and who carries it out). For larger incidents the Dutch GRIP structure applies (Gecoördineerde Regionale Incidentbestrijdingsprocedure, the coordinated regional incident response procedure), with escalation levels from GRIP 1 to GRIP 5 determining which layers of government and which crisis teams are activated as an incident grows larger or more complex.

Internationally, the CERC model (Crisis and Emergency Risk Communication) from the US Centers for Disease Control and Prevention is widely used. It sets out six principles for crisis communication: be first (communicate quickly, before others fill in the story), be right (share only verified information), be credible (be honest, including about uncertainty), express empathy (acknowledge what people are going through), promote action (give people something to do) and show respect (treat your audience as an equal). Alongside it, the Situational Crisis Communication Theory (SCCT) developed by W. Timothy Coombs is a common framework: it links the appropriate response strategy to the type of crisis and to how far an organisation is held responsible.

AI mainly bites on the first phase of BOB: gathering signals faster, forming a first picture of what is going on more quickly. Judgement and decision-making, where trade-offs and accountability come together, remain human work. Keeping that distinction clear is essential to putting AI in the right place.

Why this matters now

The market for AI-driven detection of crisis signals on social media is growing fast: market research values it globally at 1.62 billion US dollars in 2024, rising to a projected 8.39 billion by 2032, an annual growth rate of 22.8 per cent (source: AI-Powered Social Media Crisis Detection Market, SNS Insider, 2025). That growth is visible in the Netherlands too: municipalities and safety regions are experimenting with AI-supported monitoring and draft messaging to respond faster to incidents. At the same time the risks grow with it. Disinformation and deepfakes can spread quickly during a crisis and cloud the picture. AI language models can hallucinate: state invented facts with complete confidence, which in a crisis context can cause immediate harm. And with the EU AI Act, European regulation is arriving that sets limits on the use of AI systems, including in emergency response and crisis management.

That combination, growing capability alongside growing risk, is exactly why an independent knowledge platform like this one is useful: it helps communication professionals and crisis teams deploy AI deliberately, instead of avoiding it out of fear or trusting it blindly.

What stays human work

The core of responsible AI crisis communication is the principle of human in the loop: AI supports, people decide and publish. An AI system can propose a draft message, but a communication adviser judges whether the tone is right, whether the facts hold up and whether the message fits the phase the crisis is in. A chatbot can handle frequently asked questions, but a person stays responsible for escalating as soon as a question becomes too complex or too personal. And for every decision with real impact, from an emergency alert to closing a building, accountability sits with the crisis organisation, not with an algorithm.

That makes AI in crisis communication no substitute for craft, but an acceleration of the preparatory work, so people have more time for the judgements that genuinely matter.

CrisisMaatje connects the AI you already use to a crisis platform. There is no new application to learn: you work in Claude, ChatGPT or Copilot and ask for help in plain language. Your AI opens the crisis in CrisisRadar, records the first facts and the timeline, starts the clocks for statutory reporting deadlines and drafts a message. You review it, adjust it and decide whether it goes out.

That is exactly the division of labour this platform keeps coming back to: AI speeds up the preparatory work, people keep control and accountability. What sets it apart from a standalone AI chat is that everything is recorded in one place, where colleagues can follow along and carry on working, on a computer or a phone. It also leaves you with a file afterwards, so you can account for what you did and when.

You can start with a practice crisis, so you can explore it calmly before anything is at stake. Connecting is free and needs no credit card; the CrisisRadar account is created as you connect.