How Should a Company Mark an Old Crisis as Resolved for AI Systems?
To mark an old crisis as resolved for AI systems, publish a factual, dated account of what happened, what you did and how the matter was closed. Independent confirmation is the strongest part: a regulator's decision, an audit result, a court outcome or a passed inspection. Update your original statements rather than deleting them, so the record shows the resolution. Ask publishers whose articles are cited to add follow-up notes, and earn fresh coverage of the fix. AI assistants keep surfacing old crises because the coverage at the time was heavy and the resolution was quiet. Often the only sources describe the problem, not how it ended. Your task is to make the ending as easy to find as the beginning, without hiding or rewriting what happened.
Why do old crises keep appearing in AI answers?
- Coverage is lopsided. A breach, recall or scandal generates many articles. Its resolution rarely does.
- Companies delete their statements, removing the only source that described the fix.
- Articles are never updated, so they still read as current.
- Questions invite it. "Is [company] safe?" or "Has [company] had problems?" pull in the most prominent negative coverage.
- Model memory holds the event from the period when coverage peaked.
What should a resolution page contain?
| Element | Example |
|---|---|
| What happened | "In March 2025, a configuration error exposed customer email addresses." |
| When and how it was found | "We identified it on 12 March and closed the exposure the same day." |
| What you did | "We notified affected customers and the regulator, engaged an independent security firm and rebuilt our access controls." |
| How it was closed | "The regulator closed its inquiry in October 2025 with no further action." |
| Independent confirmation | "Our SOC 2 Type II report for 2026 covers the new controls." |
| What is different now | Specific changes to processes, people or products |
| Date of the update | "Updated February 2026" |
Keep the tone factual and plain. Avoid minimising language such as "a minor incident", which reads as spin to people and assistants alike. Link the page from your about, trust or security page, not only from a news archive.
How should you handle your original statements?
Update them, do not delete them. Add a dated note at the top: "Update, October 2025: this matter is resolved. The regulator closed its inquiry with no further action. Read our full account." Deleting the original statement leaves third-party coverage as the only record of the crisis, without your update. Why deleting pages can backfire more generally is covered in whether deleting old brand pages can make Gemini think the company disappeared.
How should you deal with third-party coverage?
- Do not ask for corrections of accurate reporting. The articles were right at the time.
- Ask for an update note where the resolution is significant and documented: "Update: the regulator closed its inquiry in October 2025."
- Offer a follow-up story, where there is genuine news in what changed, such as new safety measures or an independent audit.
- Update your profiles and databases where they mention the crisis as ongoing.
- Respond to reviews that refer to the crisis with a short factual note on what changed.
What should a restaurant publish when Gemini summarizes a food-safety closure?
A food-safety closure is a common case, and a sensitive one, because diners ask "Is [restaurant] safe to eat at?" long after the problem is fixed. Many local health departments publish inspection results online, and those records carry weight. Publish:
- A short, dated note on your website: when the closure happened, why, what you changed and when you reopened.
- The result of your re-inspection, with the date and any grade or score, and a link to the health department's public record.
- What changed, such as staff training, new equipment, a new supplier or a new manager, stated specifically.
- Your current inspection status, kept up to date after each routine inspection.
Also update your Business Profile description or post an update about the reopening. Respond to reviews that mention the closure with a factual line and the re-inspection result. Ask local media that covered the closure whether they will note the reopening. Then test "Is [restaurant] safe?", "Was [restaurant] shut down?" and "[restaurant] health inspection" in Gemini and Google's AI features.
What should you avoid?
- Rewriting history. Editing old statements to remove what happened is easy to spot, and it undermines trust in the resolution.
- Flooding the web with positive content to bury the story. Thin, repetitive pages rarely change AI answers and can break search engines' spam policies.
- Paying for suppression services that promise to remove accurate coverage. How to assess such promises is covered in how to audit a reputation agency's promise to suppress negative AI answers.
- Legal threats against accurate reporting, which often generate new coverage of the old problem.
- Fake reviews or testimonials to offset complaints. They are prohibited by consumer protection rules in many countries.
The credible route is slower but durable: a clear record, independent confirmation and genuine evidence of change.
Does fact-check markup help?
Not in Google Search. In 2025 Google announced it is phasing out ClaimReview, the markup used for fact-check results, in Search; it remains supported only in Google's Fact Check Explorer tool, which is built for fact-checking publishers. Put the effort into clear visible text, dated updates and independent confirmation. Whether a fact-check landing page can correct AI answers is covered in whether a landing page with FactCheck schema can correct outdated AI answers.
How do you test whether answers reflect the resolution?
Ask ChatGPT, Gemini, Perplexity and Claude, several times each:
- "Has [company] had any problems?"
- "Is [company] safe to use?"
- "What happened with [company]'s [crisis]?"
- "Is [crisis] still ongoing?"
Record whether answers mention the resolution, whether they describe the crisis as current, and which sources they cite. Aim for answers that describe the event and its resolution accurately. Complete silence about a significant past event is unlikely, and it is not the right goal.
What might an illustrative case look like?
Consider an illustrative software company whose 2024 data exposure still dominated AI answers in 2026. Assistants described it as "facing a regulatory investigation over a data leak". The company had removed its incident statement a year earlier, and the regulator's closure had never been reported. It restored the statement with a dated update and published a resolution page with the timeline, the regulator's closure and its new security audit. It asked two publishers to add update notes, and one also ran a short follow-up on the new controls. Three months later, most answers described the exposure as resolved, citing the resolution page and the updated articles.
What do people ask about old crises and AI answers?
Can we get old crisis articles removed?
Only in rare cases, such as inaccurate or unlawful content. Accurate reporting stays. Aim for updates and a clear resolution record.
Should the resolution page be prominent?
Findable rather than prominent: linked from your about or trust page, and titled clearly.
How long do old crises keep appearing?
Search-based answers can reflect a clear resolution within weeks. Model memory can hold the event longer, which is why the resolution record must stay online.
Who should approve the resolution page?
Legal or compliance, communications and the executive responsible for the fix. Agree the wording once, then reuse it in replies, profiles and briefings so every source says the same thing.
Published by The Kovmere Team · updated Sep 29, 2026 · More resources