A reputation crisis no longer takes days to build. It takes minutes, and it keeps running long after the news cycle moves on.
Key takeaways
- A crisis runs on three clocks: the platform clock (minutes), the stakeholder clock (hours to days) and the AI memory clock (months).
- The decisions that decide the outcome are made before the crisis: who pauses, who speaks and who protects the person at the centre.
- Your next crisis may not start with you. Sponsors, partners and your wider category can pull your brand in.
- AI search keeps repeating a story long after the news moves on, so monitor what ChatGPT, Gemini and Google’s AI answers say about you.
- Tools can monitor and draft at speed. A named person must still decide and approve.
Why most crisis plans fail
Most crisis plans were written for a world where a story broke, journalists called and a spokesperson had a day to respond. That world has gone. A single video can cross platforms, languages and time zones before your legal team finishes reading the first email.
Plans fail for the same three reasons. They assume the crisis starts with you. They assume it stays on one platform. And they assume it ends when the headlines do. None of that holds in 2026.
The three clocks of a crisis
Think of a crisis as three clocks ticking at once. Most brands only watch the first.
| Clock | Pace | What happens | The question to ask |
|---|---|---|---|
| Platform | Minutes | The story spreads and jumps between platforms | Where would we see this first, and who is watching? |
| Stakeholder | Hours to days | Sponsors, partners, customers, regulators and staff react | Who needs to hear from us, and in what order? |
| AI memory | Months | AI answers keep summarising the story after the headlines fade | What do ChatGPT, Gemini and Google’s AI answers say about us this week? |
Most crisis plans stop at clock two. Clock three is where the damage lingers, and it is the one most teams do not know exists.
Run a simple test. Ask three AI tools: “What happened with [your brand]?”, “Is [your brand] trustworthy?” and “Should I buy from [your brand]?” Note what is said, which sources are cited and how long it lasts.
Case 1: Hyrox Beijing and the decision gap
On 12 September, during the women’s Pro race at Hyrox Beijing, an elite athlete suffered severe gastrointestinal distress, kept racing and finished first. Footage spread on Chinese social media. By Monday her identity had been revealed online and she had received threats of violence. Hyrox issued a statement that anyone threatening or encouraging harm would be permanently banned (Malay Mail).
Days later, a Hyrox co-founder said the company had made a mistake by not reacting immediately, and its procedures changed. The athlete apologised and withdrew retroactively, forfeiting her points. A sponsor’s collaboration collection also appears to have been removed (Newsweek).
I have left the athlete unnamed on purpose. She received threats, and the lesson here is about the organisation’s decisions. In my reading of the public record, three decisions had not been made in advance:
- Who pauses the race? Other competitors shared the course, and one publicly asked Hyrox to take responsibility, including refunds or compensation.
- Who speaks, and when? Hyrox China posted a statement the same day, but the admission that mattered came days later.
- Who protects the person? Her identity surfaced and threats followed before the organiser’s statement against abuse.
None of these needs AI. They need someone to have decided in advance. The gap between the race on Saturday and the statement protecting the athlete on Monday was two days. In the viral era, that is a very long time.
Case 2: Wang Wang, a story with no brand at the centre
This case involves animal cruelty, so I will keep the description brief.
In Jieyang, Guangdong, videos circulated of four boys under 14 abusing a stray mother dog and her puppies to death. Authorities confirmed the case on 30 June and sent the boys to specialised education facilities (Mothership). Public anger then grew. A petition reached nearly 80,000 signatures in two weeks by the organiser’s own count, and international animal welfare groups spoke out (Change.org). In Singapore, supporters put up a billboard in Selegie, organised through a Facebook group (Mothership).
No brand sat at the centre of this story. It still crossed platforms, languages and, eventually, the street.
Reports also said an official appeal to stop sharing information had the opposite effect (Free Press Journal). Three lessons follow for any brand:
- The origin is rarely yours. Listen to your category, partners and sponsored events, not only your own channels.
- Stories outrun dashboards. Watch for the jump from public posts to private groups, messaging apps and the street.
- Containment extends the story. Treat visible suppression as a risk, not a plan.
Case 3: The bystander problem
A crisis does not stop at the brand in the middle of it. Sponsors and partners get pulled in, and their silence is read as a message.
While Hyrox was in crisis in Beijing, we tracked how visible a Singapore sponsor of the event stayed across social media and AI search, using Meltwater for social and BrightEdge for AEO and GEO visibility. Their total share of voice and AI brand visibility dipped month-on-month during the same period. While it’s an observation and not necessarily intended, it could be a deliberate approach at their end to stay quiet during this period.
Silence can be a smart strategy. It avoids feeding the story, protects a partner relationship and buys time for facts. It can also be the absence of a plan. The difference is that a decision has an owner, a trigger and a time to review it.
So ask yourself: if a partner you sponsor is in crisis tomorrow, would your silence be a decision?
What to put in place before the crisis
Every decision below can be made in calm, before you need it.
- Authority. Who can pause, who speaks and who signs off, ideally within 30 minutes.
- Holding statements. Pre-approved drafts for your top five risks, including a partner or sponsor in crisis.
- Listening. Social, news and AI search, in every language your audience uses.
- Partner protocols. Agreed triggers for when a sponsor speaks, pauses or distances.
- People first. A plan for the person at the centre, including threats and pile-ons.
The first hour
This is a working standard, not research, so adapt the timings to your sector and risk.
- 0 to 15 minutes, detect and verify. Is it real, is it spreading and who is it about?
- 15 to 30 minutes, decide. Pause, speak or watch, with one owner for each call.
- 30 to 60 minutes, say something. A holding statement, a named spokesperson and a time for the next update.
The time of the next update matters most. The silence between updates is where speculation grows.
AI tools that help you respond, not just monitor
Detection used to be the hard part. Now the hard part is responding at speed without losing control. A practical stack has four layers, and smaller teams can cover the first three without an enterprise budget.
| Layer | Enterprise | Smaller teams | The job |
|---|---|---|---|
| Detect | Meltwater Mira Studio | Brand24, Mention, Google Alerts | Spot spikes and shifts in sentiment |
| See AI answers | BrightEdge | Otterly.ai, Peec AI, Semrush or Ahrefs add-ons | See what ChatGPT, Gemini and AI Overviews say |
| Respond | Airyn (OneByZero), Sprinklr, Sprout Social | NapoleonCat, replient.ai, Hootsuite | Draft, route and escalate replies |
| Govern | Named approver, kill switch, weekly review | The same controls | Keep a person accountable |
Response automation can be powerful. PLDT and Smart in the Philippines report that Airyn automates 93.7% of their public social replies and lifted engagement 2.6 times, according to the vendor’s case study. Treat those as vendor-reported results for a customer service use case, not a crisis one.
Which is the point. The more you automate the 93.7%, the more your reputation depends on the other 6.3%. Four controls keep that 6.3% safe:
- A kill switch that pauses every auto-reply the moment a spike is detected.
- Escalation triggers for toxicity, volume spikes, media or VIP mentions, and anything legal or safety related.
- Human sign-off. AI drafts, and a named person approves every public statement.
- A weekly review of what the bot got wrong.
FAQs
What are the three clocks of a crisis?
The platform clock (minutes), the stakeholder clock (hours to days) and the AI memory clock (months). Most crisis plans manage only the first.
What should a company do in the first hour of a social media crisis?
Spend 15 minutes detecting and verifying, 15 minutes deciding who pauses, speaks or watches, and the next 30 minutes issuing a holding statement with a named spokesperson and a time for the next update.
How do you check what AI says about your brand after a crisis?
Ask ChatGPT, Gemini, Perplexity and Google AI Overviews what happened, whether you can be trusted and whether to buy, then track what is said, which sources are cited and how long it lasts.
Should crisis responses be automated?
Automate routine replies, never public statements. Keep a kill switch, escalation triggers and a named person who approves every statement.
Work with MAMC
Mad About Marketing Consulting advises boards and leadership teams of private and scaling companies on marketing, customer experience and AI transformation, including reputation and crisis readiness. MAMC was named Consultancy Firm of the Year by IE100 Magazine. Our founder brings 26 years in marketing across financial services, consulting and healthcare in APAC, including crisis communications preparation, spokesperson training and sponsorship oversight.
To discuss how MAMC can support your board, view our services here.
