Prompt of the Week: The Crisis Afterlife Check

Hi Reader,

Remember the issue you narrowly avoided a few years ago?

You got the statement out. Social didn't spike. The coverage faded. And you moved on.

AI did not.

This week, Sefiani (part of Clarity Global) tested what AI assistants say about brands with data breaches that were resolved years ago. Ask a broad category question and the breach barely comes up (2.7% of answers). Ask "can I trust this brand?" and it surfaces in 42.2% of answers. For one organization, 59.1%.

Think of it like a house that flooded in 2019. New drywall, new wiring, fully fixed. But the inspection report is still in the county file, and every buyer's agent pulls it first.

Here is the part that should get your attention. Owned content, like your own website, showed up in almost 90% of those AI answers. As Mandy Galmes from Sefiani put it, if you skip the step of publishing a clear account, "AI will build the story and define your reputation for you."

So the question is simple. What does AI say about the crises you averted, and who wrote that version?

👇🏽 Here's the prompt:

🎯 Expert Mode: The "Crisis Afterlife Check"

Scenario: You closed out a crisis, breach, recall, or scandal years ago, and you assume it is history. You need to find out what AI assistants say about it today, spot what is wrong or outdated, and publish the facts before a prospect, reporter, or investor reads someone else's version.

Instructions:

1️⃣ Pick one resolved incident. Write down the verified facts in one document: what happened, when it was resolved, who was affected, what you changed, and what is different today. Keep it plain and factual.

2️⃣ Open ChatGPT, Gemini, and Claude. Ask each one: "Is [Your Brand] trustworthy?" and "Has [Your Brand] had any controversies or incidents?" Paste every answer into the same document, along with any sources the tool cites.

3️⃣ Paste the whole document into Claude with the prompt below.

Act as a crisis communications strategist who specializes in how AI systems describe a company's past incidents. Below are two things: (1) the verified facts about [INCIDENT], which was resolved on [DATE], and (2) what three AI assistants currently say when asked whether [BRAND] can be trusted.

First, compare each AI answer to the verified facts and label every claim as accurate, outdated, missing context, or wrong.

Second, identify which sources the AI answers appear to rely on, and whether each one is owned, earned, or third-party.

Third, rate the overall reputational drag on a scale of 1 to 5 and explain your rating in two sentences.

Fourth, draft a plain, factual one-page account for our website that covers what happened, who was affected, what we changed, and what we do today. Write it so an AI system can quote it accurately.

Fifth, give me five questions I can rerun every month to track whether the story is shifting.

Do not minimize the incident or invent facts. If something is not in my verified facts, flag it as a gap.

4️⃣ Read the labels with one filter: would a skeptical customer or reporter find this answer fair? Mark every "wrong" and "outdated" claim. Those are your priorities.

5️⃣ Get legal to review the one-page account, publish it on your own site, and set a calendar reminder to rerun the five questions in 30 days. Compare the answers.

Suggested Tool: Claude (strongest at checking claims against a source document without filling in gaps)

Why You'll Love It

It shows you what a buyer, reporter, or investor sees when they ask AI about your worst week, before they bring it up in a meeting.
It gives you a factual, quotable account you control, instead of leaving the story to whatever third-party commentary AI finds first.
It turns a vague worry about "AI and our reputation" into a monthly check you can run in ten minutes.

You cannot delete the past. But you can make sure the version AI reads is the accurate one, and that it comes from you.

Talk soon,

Chris

⚡ The Cowork Version

Same check, except Claude reads your incident files and the AI answers you saved, scores them against the facts, and drops a finished kit into your folder.

Setup (one time):

Create a folder called "Crisis Afterlife" and add a "verified facts" document for one resolved incident (what happened, resolution date, who was affected, what changed, what is different today).
Ask ChatGPT, Gemini, and Claude whether your brand is trustworthy and whether it has had controversies. Save each answer as its own file in the folder, with any cited sources.
Add your current online newsroom or "trust" page text, if you have one, so Claude can see what your owned content already says.
Optional: If Claude in Chrome or another connected tool is available, let Claude run the questions itself and save the answers for you.

Run this Cowork command:

Run a Crisis Afterlife Check using the files in this folder.

1️⃣ Read the verified facts document and build a short fact sheet: what happened, resolution date, who was affected, what changed, and what is different today.

2️⃣ Read each saved AI answer. Label every claim about the incident as accurate, outdated, missing context, or wrong, and quote the exact line you are labeling.

3️⃣ List the sources each AI answer cites or appears to rely on. Tag each one as owned, earned, or third party.

4️⃣ Compare our current website content to the fact sheet and flag anything missing, vague, or out of date.

5️⃣ Rate the overall reputational drag from 1 to 5 and explain it in two sentences. Do not invent facts. Flag any gap in my files instead of guessing.

6️⃣ Draft a plain, factual one-page account for our website, written so an AI system can quote it accurately. Then write five questions I can rerun every month.

7️⃣ Save the results in a new folder called "Crisis Afterlife Output": a scorecard spreadsheet, the one-page account as a Word document, and a one-page summary for leadership.

What you'll get:

A claim-by-claim scorecard showing what each AI assistant gets right, wrong, or leaves out
A source map showing whether the story is coming from your owned content, earned media, or third parties
A draft one-page factual account, ready for legal review and your website
A five-question monthly tracker so you can see whether the story shifts
A one-page leadership summary you can forward without reformatting

Want it on autopilot? Ask Claude to save this as a skill, or to schedule it monthly against fresh AI answers. You get a standing reputation check instead of a one-time scramble.

Here are a few articles from the past week to help keep you up to speed:

1. AI is giving corporate crises a second life, long after the headlines fade

Marketing-Interactive

October 29, 2026

Sefiani's research found that AI assistants resurface resolved data breaches most often when users ask whether a specific brand can be trusted. The old fade-out that let brands rebuild is disappearing. This is the story behind this week's prompt, and it makes the case for publishing your own factual account before a model builds one for you.

2. AI Influence, Not Control: Five Takeaways From Muck Rack's 2026 Generative Pulse Summit

PRNEWS

November 2, 2026

You cannot control what AI says about your brand, but you can shape the sources it draws from. The summit advice is practical: audit your own pages, track a set of key prompts across several models, and add AI monitoring to your crisis plan with a named owner. If your crisis plan does not mention AI yet, this is your checklist.

3. PR Fumbled SEO. Don't Fumble AI Visibility.

Spin Sucks

November 3, 2026

Gini Dietrich argues that PR gave away SEO and cannot afford to give away AI visibility too. She argues that AI models learn heavily from earned and owned media, which is already the PR toolkit. It is a useful frame for your 2027 budget conversation: get a baseline, then measure whether you are absent, mentioned, cited, or featured.

4. The Missing Half of AI Measurement

Institute for Public Relations

November 4, 2026

Jamin Spitzer warns that cost and speed are crowding out real insight in AI-driven communications measurement. If leadership asks you to prove the value of AI this quarter, this is a good counterweight. Efficiency numbers are easy to report, and they are only half the story.

5. AI Data Centers Have a PR Problem. OpenAI Is Proposing a Rebrand

CNET

November 3, 2026

OpenAI is reportedly suggesting that data centers be renamed as part of a marketing push after the midterms. It is a live test of a question every comms team gets eventually: can a new name fix a trust problem? Watch how the coverage responds, because the answer usually depends on whether anything changed behind the label.

Ready to Future-Proof Your PR Strategy with AI? Here’s How I Can Help:

AI Training & Implementation Workshops:
Hands-on, live sessions that guide your team through integrating AI into their workflows and content creation, building practical skills that matter now.
AI Strategy & Implementation Sessions (1:1):
Personalized, one-on-one sessions to uncover your biggest AI opportunities, whether it’s automation, adoption, or designing a tailored roadmap for success.
AI Advisory & Consulting (Retainer or Project-Based):
Ongoing support to embed AI tools into your core processes, automate key workflows and future-proof your communications strategy with measurable impact.