How to Manage a Brand Crisis When AI Assistants Repeat It

During a crisis, AI assistants summarise what the web says about your brand, often within days. You cannot edit their answers directly, but you can publish a clear, factual statement, correct errors at the sources AI reads, and keep updating those sources until the answers catch up.

Why AI Answers Matter in a Crisis

Customers, journalists and job candidates now ask questions like “Is [brand] safe to use after the data breach?” or “What happened with [brand]’s delivery delays?” in a chat window instead of a search box. The assistant gives one confident summary. If that summary is built on the first wave of angry posts and early reports, it may describe a problem you have already fixed, or mix up facts.

Assistants that browse the web, such as ChatGPT with search, Perplexity and Copilot, can pick up new pages quickly. Answers that rely mostly on training data change slowly, because a model only learns from newer data when it is updated. That is why a crisis plan for AI has a short-term part and a long-term part.

The First 48 Hours

1. Check what assistants say. Ask five or six real questions in each major assistant: “What happened with [brand]?”, “Should I still buy from [brand]?”, “[brand] refund policy after the outage”. Save the answers and the sources they cite.
2. Publish one official statement page. Put it on your own site at a stable URL, with a date, a plain summary of what happened, what you are doing, and what customers should do. Avoid PDFs or images of text; AI tools read normal HTML best.
3. Make sure crawlers can reach it. Do not block the page in robots.txt and link to it from your homepage or news section so it is found fast.
4. Brief your spokespeople. Every quote you give the press becomes source material. Keep the same core facts in every interview.

Correcting Errors at the Source

If an assistant gets something wrong, look at where the error came from. Perplexity and Copilot show their sources, which makes this easier. Typical fixes:

• Ask a publication to correct a factual mistake, with evidence. Most reputable outlets have a corrections process.
• Update your own FAQ, help centre and social profiles so they all state the same facts.
• Answer customer reviews calmly and with specifics. Reviews on major platforms are often read by AI tools.
• Use the feedback buttons inside the assistants to flag inaccurate answers. It is not a guaranteed fix, but it is a legitimate channel.

For broader errors that outlast the crisis, see our guide on fixing AI hallucinations about your brand.

What Not to Do

Do not flood forums with fake positive posts or buy reviews to drown out criticism. It is easy to spot, it breaks platform rules, and if it is exposed it becomes a second crisis that AI tools will happily summarise. Do not delete your old statements either; update them with a dated note so the timeline stays clear.

The Weeks After

Once the story calms down, publish a follow-up: what changed, what you learned, what customers can expect now. This gives assistants a newer, more complete source than the early headlines. Keep asking the same questions every week and note whether answers now mention the resolution. Our article on brand reputation in AI chatbots covers the longer-term work.

How Keboost Helps

Keboost tracks the prompts people ask about your brand across ChatGPT, Gemini, Claude, Perplexity and Copilot, so you can see how answers change during and after an incident and which sources they lean on. It also helps you create and distribute clear, factual content so assistants have accurate material to find. We cannot remove a story from AI answers or promise a date when they change, but we can help you give them better sources. Run a free AI visibility check to see what assistants say about you today.