AI is rewiring the reputation playbook, in a crisis and beyond
AI‑mediated channels such as Large Language Models (LLMs) have rapidly become a critical part of the reputation landscape, sitting alongside traditional, digital, and social channels.
Virtually overnight, LLMs, AI‑enhanced search, and AI‑driven aggregators have become primary interfaces through which people can access, interpret, and evaluate information.
While the digital age has seen several transformative shifts, including the rise of social media and the 24-hour news cycle, AI systems introduce a fundamentally new dynamic to organisational reputation because of the way they retrieve, summarise, and frame information.
LLMs, in particular, affect reputation by playing three interconnected roles:
Information source: giving people fast, conversational access to large volumes of content.
Representation system: summarising, reframing, and connecting information in ways that influence how issues appear.
Analysis engine: ranking, prioritising, and filtering information, influencing perceptions of what appears important or credible.
In the simplest terms, LLMs influence what information users see, how it is presented, and how it is interpreted.
Protecting reputation in a new world
Organisations must extend internal capability beyond 'traditional' channels and build new capability in AI-mediated environments, because, while distilling overwhelming amounts of content at speed may be incredibly useful, it is also where things can go very wrong when it comes to reputation.
One challenge, arising from the information used for training and/or the sources retrieved and summarised, is that AI-powered tools can oversimplify complex information, distort timelines and amplify misleading, low‑quality or outdated content. They also have the potential to shape how organisations, issues and events are understood. For example, they can blend verified and unverified information in ways that make sources appear equally credible, making it harder to distinguish fact from fiction.
Another challenge comes from the persistence of information across the AI ecosystem. Whether embedded in training data, archived online content, search indexes, or AI-generated summaries, information can remain discoverable and can be resurfaced long after public attention has moved on. Old issues can appear current, resolved issues can seem unresolved, minor issues can look major, rumours can surface as facts, and narratives can be reframed without context or correction.
The flow-on effect of flawed information
The public's ongoing engagement with unreliable and outdated information continues to accelerate and intensify, and this is influencing how issues and crises are being interpreted and perceived by audiences both in real-time, and long after events have passed.
And it won’t always be isolated to a small group. Thanks to the psychological tendency of ‘guilty by association’, the consequence of any crisis has the potential to extend beyond the organisation(s) directly involved, even when no evidence exists. AI-powered search, recommendation and conversational systems can hasten and intensify these conditions by surfacing old, decontextualised, or misleading associations and connections, making the implications farther‑reaching and longer-lasting.
In this supercharged environment of information immediacy, another threat has emerged, highlighted by the Centre for Emerging Technology and Security (CETaS) in the UK.
CETaS outlines how a polluted online information ecosystem, driven by the adversarial use of AI, can escalate crises dramatically and contribute to real-world harm. Drawing on recent events, the paper demonstrates how malicious actors can weaponise AI tools to create misleading content and amplify harmful narratives. This includes poisoning LLM training data, exploiting AI‑driven news aggregators to spread misinformation at scale, and manipulating how AI 'chatbots' (LLMs) and social media platforms deliver personalised information in real-time.
AI tools are also increasingly being used to organise and amplify complaints and grievances against organisations and industries through coordinated campaigns.
While coordinated complaint behaviour itself isn’t new, the speed and scale at which it can now occur is unprecedented. Often originating from real users’ accounts, these tools can rapidly generate emails and social media posts that appear entirely unique by varying the wording, tone, formatting and length, while delivering the same coordinated message. This makes orchestrated activity significantly harder to detect.
The motivations behind this behaviour aren’t always malicious, but the effect is the same: AI‑generated activity blurs the line between genuine community sentiment and strategically engineered feedback. So, for organisations trying to understand public perception, the task of distinguishing authentic concerns from artificial ones is increasingly difficult.
AI risk and reward
While it is important to acknowledge these risks and develop the capability to manage them, this does not mean that AI-powered tools are inherently negative. The challenge for organisations is that the systems creating new pathways for misinformation, distortion and reputational harm are the same systems creating new opportunities for accessibility, discoverability, transparency and understanding.
The new Corporate Affairs playbook
AI changes the speed, scale, and dynamics of trust and reputation, and that means organisations can't rely on the old playbook. The fundamentals still matter, but they’re no longer solely sufficient. New methods are needed to keep pace.
There are nine areas where communications and corporate affairs teams should be focusing right now:
Governance: Advocating for organisational readiness, from policies for AI use and monitoring to governance and frameworks around misinformation response protocols.
Channel management: Treating AI-mediated channels, including LLMs, AI search tools and AI-driven aggregators, as reputation channels, and monitoring them accordingly – tracking how they summarise organisational information, identifying what is being resurfaced, monitoring how they frame connections and associations, and comparing outputs across models.
Risk management: Building an AI information risk early‑warning system, detecting coordinated messaging patterns, misinformation, rumours and outdated information appearing as current facts.
Crisis preparedness: Preparing for adversarial AI use, coordinating with cybersecurity and risk teams to scenario plan and developing rapid response protocols alongside wider crisis preparedness activities.
Upskilling and reskilling: Building team literacy for the new information environment, across AI, LLMs, and evolving crisis dynamics.
Stakeholder management: Strengthening direct stakeholder channels and investing in trusted, authentic interactions.
Information stewardship: Building and maintaining consistent, factual and easily discoverable sources of organisational information, with clear timelines, authoritative reference points and accessible context that AI systems can retrieve and rely upon. When issues arise, supporting context restoration through consistent factual anchor points and clear records of events.
Analysis: Separating genuine sentiment from synthetic, or AI‑amplified noise by analysing source diversity, temporal anomalies, linguistic fingerprints and human validation.
Tracking: Re‑evaluating what ‘community sentiment’ actually means in a part real, part amplified, part synthetic, part algorithmically-shaped environment.
