Reputation crisis: what to do after the first week
A reputation crisis changes after week one. Here is the plan for weeks two to four: when to respond, what to correct, how to monitor and when to stop.
After the first week of a reputation crisis, the job changes from stopping the spread to deciding what still deserves action. In weeks two to four you respond only when there is a new fact, a new claim or a clear error, you pursue removal or correction only where a route exists, you lower monitoring to a cadence you can keep up, and you stop engaging with anything that only repeats the story.
Most guides stop at the first 48 hours or jump to a year-long recovery. The weeks in between are where avoidable damage happens: a second statement that reopens the story, a removal attempt that becomes its own article, a stakeholder who gets three answers from three people, a monitoring routine that burns out the one person running it.
This article covers only that phase and assumes the first-week work is done.
- Weeks two to four are for decisions, not volume. Act on new facts, clear errors and unlawful content, and leave repetition alone.
- Respond again only when something new has been added to the story or the record is wrong in a way you can prove. Restating your position is not new information.
- Move monitoring to a fixed rhythm with named escalation triggers.
- Give stakeholders one updated account on a schedule, not a reply to each question as it arrives.
- Define in advance what "done engaging" looks like, then hand the file to a longer plan.
Weeks two to four are a different job from the first week
The first week is containment: speed, facts, a holding statement. Weeks two to four are judgment: the story has either found its shape or is still growing, and your job is to tell which.
Three things change at once. Coverage that was constant becomes occasional. Your capacity drops, because the people who handled week one are tired. And the audience narrows to people with a reason to care: customers, partners, employers, sponsors, lenders, regulators and journalists doing a follow-up. You are managing the people who will make a decision about you, and what they see when they check.
What the first week should have left you
Confirm you hold five things, and build any missing one in week two: a fact base (established, disputed, unknown, each tied to evidence), a log of every item about you, a holding position, a named decision owner with a deputy, and a record of who has been told what.
Related readCrisis Communication to Employees: Who to Tell FirstWeek-by-week plan for weeks two to four
Each week has one main job, one decision to settle and one thing to stop doing. A story with a legal process, a regulator or an investigation has its own calendar, and that calendar overrides this table.
| Week | Main job | Decision to settle | Stop doing |
|---|---|---|---|
| Week 2 | Triage the log: sort every item into accurate, inaccurate, unlawful or opinion | Which items get a correction or removal request, and which are left alone | Replying to individual comments and posts |
| Week 3 | Run the requests and the stakeholder updates; lower monitoring to a fixed rhythm | Whether any new development justifies a second public statement | Checking every few hours; drafting statements nobody asked for |
| Week 4 | Review results: what moved, what was refused, what remains in search and AI answers | What the exit test says: close the crisis phase or keep it open one more cycle | Treating the file as an emergency |
When to respond again and when to stay silent
Respond again only when at least one of three things is true: there is a new fact, a new claim, or a clear error you can document. If none is true, silence is the correct response, and it takes discipline.
A new fact is a decision, a finding, a correction by the publisher, or a court or regulator step. A new claim is an allegation not in the earlier coverage. A clear error is one a neutral reader can check, such as a date, role, amount or name.
Repeated criticism and opinion are not triggers.
| What you see | Respond? | How | Risk if you do it wrong |
|---|---|---|---|
| Same claim, new outlet, no new facts | No | Log it; check it is accurate | A second statement signals the story is still alive |
| A new allegation | Yes, if it is false or you can answer it | One short factual reply, to the source or the people affected | Silence lets the new claim become the version people repeat |
| A factual error in coverage | Yes | Private correction request with evidence, to the publisher | A public fight over a small error becomes the headline |
| A real development in your favor | Yes, briefly | Update your own page and tell stakeholders first | Over-claiming a result that is not final |
| Opinion, mockery, commentary | No | Do not engage | Engagement extends the life of the thread |
| Unlawful content (for example threats or private data) | Yes, through the platform or counsel | Report; keep a record | Public argument instead of a report |
The one-reply rule
When a response is justified, write one reply, make it short, and put it where the affected audience will find it: your own page, a direct message to the people concerned, or a note to the journalist. A long rebuttal repeats every allegation and hands search engines and AI assistants a new page that contains them.
Removal or correction: which route to pursue now
Pursue removal only for content that is unlawful, private data or a clear breach of a platform rule, and pursue correction for inaccurate reporting. Accurate, lawful reporting about a real event is generally not removable, and asking for its removal tends to create a second story. The question for each item is not whether you dislike it but which test it can pass.
Related readProduct Recall Reputation: What Regulators Publish| Item | Route | What it can change | Main risk |
|---|---|---|---|
| News article with a factual error | Correction request to the publisher or its press regulator | A correction, a clarification or an update note | Slow; removal is usually not the remedy |
| Accurate news article | None for removal; build better pages instead | Nothing at the source | A removal attempt can be reported on |
| Personal contact data in a result | Search engine personal-information request | The result in that engine, not the page | Source page remains online |
| Post that breaks a platform rule | Platform report with evidence | The post, if the rule applies | Refusal; appeal rules vary |
| Defamatory or unlawful statement | Counsel, legal letter, court route | The statement and sometimes its search visibility | Cost, delay, publicity |
| Page already corrected at source | Search engine request to update the old result | The outdated snippet in results | Only works once the page has changed |
What the sources say about each route
Press regulators differ by country. In the United Kingdom, the Independent Press Standards Organisation (IPSO) publishes the Editors' Code, and IPSO's explanation of Clause 1 says a significant inaccuracy, misleading statement or distortion must be corrected promptly and with due prominence. It also says an inaccuracy alone is not a breach and gives a typographical error as one that would not qualify. This is a UK rule for IPSO-regulated publications, not a universal right to a correction, and the correction route in this section is UK-only.
The timing matters in a month-one plan. IPSO's time limits page says it can consider complaints made within four months of publication, and complaints about material still published online within 12 months of publication, if it is still possible to investigate and adjudicate fairly. A complaint about week-one coverage prepared in weeks two to four is still inside the four-month window, so there is no reason to rush a weak one.
For personal information, Google's Results about you page says the tool covers results showing items such as a phone number, home address, email address, or government ID numbers, and that Google may approve such requests. It also says removing a result does not delete the information from the internet, and that results on online newspapers, government or educational sites and business websites do not get a "Remove result" option. A news article about a crisis is therefore outside it. Read the page before filing, because the scope changes.
The Streisand risk, stated plainly
Every removal attempt can itself be reported. A legal threat to a small outlet or a flurry of takedown requests can become a follow-up story. Use private channels, keep the tone factual, and send only requests you can support with evidence.
Related readReputation Crisis First Hour: A Checklist in OrderHow to move monitoring from hourly to a sustainable cadence
Reduce monitoring in steps, and replace attention with triggers. If you drop from hourly checks to nothing, you will miss the next development. If you stay at hourly, you will usually stop by the end of week three without deciding to.
- Week two: every four hoursCheck named searches, social mentions and AI answers on a schedule. Log only changes, not everything you see.
- Week three: twice a dayCheck once in the morning and once in the evening. Set alerts for your name, company name and the main allegation wording, so the alerts carry the overnight load.
- Week four: once a day, plus a weekly reviewDo a short daily scan. Once a week, repeat a fixed set of searches and AI prompts and compare to the previous week.
- Write the triggers downAgree in advance what moves you back up a level: a new outlet with original reporting, a new allegation, a regulator or court step, a result entering the first page, or a wrong AI answer.
- Name the backupOne person runs the routine. A second person covers days off and takes over if the first is unavailable.
To show how much work the ladder removes, take a simple case with stated assumptions: checks follow the schedule in the chart, around the clock, and each check counts as one pass over the same search list. The figures are computed from those assumptions, not measured.
Computed from the assumptions in the caption: 24 x 7, 6 x 7, 2 x 7 and 1 x 7 passes. The schedule is an example, not a recommendation for every case.
Alerts help with the overnight load. According to the Google Alerts help page, Google Alerts can email you when new results for a topic show up in Google Search, and you can change how often you are notified. Treat alerts as a net with holes: they do not replace the weekly check of the actual results and of what AI assistants say. A managed version exists as reputation monitoring, but the routine above works without it.
Log changes only: date, where, what changed, whether it is accurate, action taken.
What to tell stakeholders who are still asking
Give stakeholders one account, updated on a schedule, and send it before they ask. In weeks two to four the questions come from lenders, partners, customers, employers, boards and sponsors, and the answer matters less than its consistency. Three answers from three people look like three stories.
Build the update from the fact base: what is established, what is open, what you are doing and when the next update comes. Name what you do not know, since an unknown stated early is less damaging than an assumption corrected later.
Related readArrested: What to Do for Your Reputation in Week One- Set one update date per audience and keep it, even when nothing has changed.
- Do not promise an outcome you do not control, such as a regulator or court decision.
- Check any legal constraint before sharing detail about a matter under proceedings. Your counsel decides what can be said.
When a stakeholder asks for a fact you cannot share, say so and why, in one sentence.
How search results and AI answers move as coverage ages
Search results and AI answers do not reset when the news cycle moves on. They change as new pages appear, old pages are updated, and ranking systems weigh what is fresh against what is established. No source we could read gives a reliable timetable for when a crisis story drops, so plan for movement you can observe rather than a date.
Google's guide to ranking systems says it has "query deserves freshness" systems that show fresher content for queries where it would be expected, such as news articles after an earthquake. It also describes original content systems that show original content, including original reporting, prominently over pages that only cite it. While a story is active, fresh and original coverage can therefore sit high.
What this means for weeks two to four:
- Expect the top of the results to shift as follow-ups appear, even when nothing you did changed.
- A correction or update note at the source helps only if it is on the page that ranks.
- Your own accurate pages need time to be crawled and weighed. They are not a week-three fix.
- Check your name alone, with the topic, and in the phrasing a cautious partner would type.
AI answers lag and borrow
AI assistants differ in how they answer. Some search the web and cite sources. OpenAI's help page for ChatGPT search says ChatGPT searches the web automatically when a question may benefit from it, and that responses using search include inline citations and a list of sources. Google's page on AI Overviews says each overview links to supporting web pages and that it can and will make mistakes, with a feedback option on every overview, as stated on its AI Overviews help page.
In our experience, an AI answer in month one often summarizes the pages it can find, so early, widely linked reports weigh heavily. Check the same fixed prompts every week, record the answer and its cited sources, and correct the sources before arguing with the answer.
Related readContent Creator Crisis Response: A Plan That Protects IncomeReporting a wrong answer can be a slow channel. In a complaint filed on 29 April 2024, the privacy group noyb said OpenAI had refused a request to correct data about a person, saying it was not possible to correct data, and had said it could filter or block data on certain prompts, as described on noyb's page. That is one complainant's account of one provider's position at that date, so check the provider's current process before relying on it. Our AI search reputation page covers the wider question of what assistants say about a name.
The traps of month one
Month one has three traps, each a decision that feels sensible when it is made: over-responding, reopening the story, and quietly expanding the claims. All three come from the wish to be finished with the matter.
Trap one: over-responding
Every reply is a new page and a new reason for a follow-up. The signs are a second and third statement and answers to every comment. The fix is the one-reply rule and a decision owner who can say no.
Trap two: reopening the story
You can restart the story yourself by announcing a removal campaign, defending a matter that had gone quiet, or deleting posts without explanation. Before any public step, ask whether it would be news if someone else did it. If yes, give it the care of a week-one statement.
Deleting your own content is a special case: do it once, for a stated reason, and keep a copy. A quiet mass deletion looks like concealment.
Trap three: quiet expansion of claims
Claims grow without anyone deciding they should. "We are reviewing the matter" becomes "we have found nothing wrong" in a stakeholder email. Control it with one source of truth: every message is checked against the fact base before it goes out. A message that goes further than your position is a new claim, and you will be held to it.
Related readExecutive Crisis Communication Plan for a Personal CrisisA worked example: weeks two to four for an invented case
The following is an invented scenario with generic parties and stated assumptions, written to show the decisions in order. It does not describe a real case.
Assumptions. A small regional services firm. A local news site published a report in week one alleging billing irregularities. One detail is wrong: the report gives the wrong year for a regulatory finding. The firm issued a holding statement in week one. Two other outlets repeated the report. A forum thread discusses the firm. The owner and a manager are the only two people handling the matter.
| When | What happens | Decision | Action |
|---|---|---|---|
| Week 2, day 1 | Log has seven items | Sort by accurate, inaccurate, unlawful, opinion | One factual error found; six items left alone |
| Week 2, day 3 | Wrong year found in the original report | Correction request, private, with the regulator's own document attached | Request sent to the news site's editor |
| Week 2, day 6 | A trade outlet adds a claim about a named client | New claim, so a response is allowed | One short factual note sent to the outlet and to the client |
| Week 3 | Forum thread revives; comments are opinion | No trigger met | Do not reply; monitoring drops to twice a day |
| Week 3, day 5 | Lender asks for an update | Send the scheduled stakeholder note | One page from the fact base, reviewed by counsel |
| Week 4 | Site publishes a correction note; the old snippet still shows | Source now changed, so a search request is possible | Ask the search engine to update the result; run the weekly check |
The firm did not answer the forum, announce a campaign, ask for the report to be taken down or publish a long defense. It made one private correction request, one factual note and one stakeholder update, and asked a search engine to refresh the result only after the source had changed, because a result cannot be updated until the page behind it is. An outlet may decline to correct, and the plan must work if it does.
When to stop engaging and hand over to the long plan
Stop engaging when the exit test is met, and write the test in week two so you are not deciding in week four while tired. The phase is over when you have stopped reacting, not when the story has vanished. Close it when all five are true:
- No new original reporting or new allegation has appeared for a full week.
- Every logged item has a recorded decision: corrected, reported, accepted or left alone.
- Stakeholders have received their latest scheduled update and no one is waiting for an answer.
- Monitoring has run at the weekly cadence for at least a week without a trigger.
- The search results and AI answers you track have been stable across two weekly checks.
If any one is false, keep the file open for one more cycle at the same cadence, not week-one intensity. If a trigger fires, move up one level of the ladder, deal with it, and step back down.
Related readI Posted Something I Regret and It Is Going ViralClosing the phase does not mean the matter is solved. The file moves to the slower work of building accurate pages, correcting what stays wrong and watching monthly.
Your next step if you are in week two now
In week two, do the triage first: sort every item into accurate, inaccurate, unlawful or opinion, and decide which few justify action. Write the exit test and the monitoring triggers the same day, and name the second person.
If the story is still moving, or you are not sure which items deserve action, a private review of the log and the search results can settle the first decisions. Reputation Partners handles these files, and the audit is the place to start.
Questions people ask
How long does the second phase of a reputation crisis last?
There is no fixed length. In our experience the second phase usually runs about three weeks, weeks two to four, but it ends when the exit test is met, not on a date. A legal process, a regulator or new coverage can restart it, and the calendar of that process then sets the pace.
Is it too late to ask for a correction after a week?
No. Correction routes run on their own clocks. In the United Kingdom, IPSO says it can consider complaints within four months of publication, or within 12 months for material still published online if it can still investigate fairly. This applies to IPSO-regulated publications only. Other countries and outlets have different rules, so check the publisher's policy and any regulator that applies.
Can I make negative coverage disappear in a month?
Accurate, lawful reporting generally cannot be removed on request, and a month is rarely enough to change what search results show. In a month you can correct errors, remove content that breaks a rule or the law, and start building accurate pages. Anyone promising more is promising something they cannot control.
What should I say to customers or partners who keep asking?
Send one short account on a schedule: what is established, what is open, what you are doing and when the next update comes. Use the same text for everyone, check it against your fact base, and say plainly what you cannot share and why.
When should a lawyer be involved?
Involve counsel when content may be defamatory or unlawful, when a legal letter is being considered, when a regulator or court is involved, or before you share detail about a matter under proceedings. Rules differ by country, and this article is information, not legal advice.
