Negative news does not disappear when it drops out of Google’s first page. It can continue to appear when someone asks ChatGPT, Perplexity or Google AI Mode about a company, brand or person.
The problem becomes more complicated when an old report is presented without an important update, or when an AI-generated answer describes an allegation as an established fact.
For a business, this is not simply an SEO problem. It is a reputation and information accuracy problem.
The first step is to identify the original source and check whether the AI summary is accurate. Then correct errors at the source, publish verifiable information and monitor how AI platforms describe the subject. There is no guaranteed way to make an AI platform remove an unfavourable but accurate report.
How do you recover your brand’s reputation in AI search?
You cannot directly control what ChatGPT, Perplexity or Google says about your brand. You can, however, improve the accuracy and availability of the information those systems may use.
Start with five actions:
- Identify the exact negative statement and the AI platform showing it.
- Open the cited sources and find where the claim originated.
- Ask the original publisher to correct demonstrable factual errors.
- Publish accurate, documented information on official channels.
- Monitor the same questions and sources across AI platforms.
The important distinction is between correcting false information and trying to hide unfavourable information. They are not the same job.
Why negative PR can appear in AI search summaries
A traditional search result usually gives a person a list of pages. The person decides which pages to open and what conclusions to draw.
An AI search experience can present a summary before the person visits any website. It may combine information from multiple sources and provide links to some of them.
Imagine a company that faced a product quality complaint two years ago. The company subsequently fixed the issue, published an update and changed its quality process. However, an old news report still appears prominently in search results.
A person asking an AI assistant whether the company is trustworthy may receive an answer that mentions the old complaint without adequately explaining what happened afterwards.
The original report may be accurate. The problem may be that the answer lacks important context.
There are three situations to separate:
| Situation | What it means | What to do |
|---|---|---|
| Incorrect information | The AI states something factually wrong | Document the error and report it through the relevant channel |
| Incomplete information | The original event is real, but the answer misses later developments | Publish verifiable updates and make them easy to find |
| Negative but accurate information | The source reports an event that genuinely happened | Address the issue transparently and demonstrate what changed |
A reputation team should not treat all three as misinformation.
Find out exactly what the AI is saying
Do not begin by publishing a press release or sending removal requests. First, establish what the problem actually is.
Test the brand across the platforms that matter to your customers.
Use questions such as:
- What are the major controversies involving [Brand]?
- Has [Brand] faced any customer complaints?
- What happened in the [specific incident] involving [Brand]?
- Has [Brand] resolved the issue?
- What do customers say about [Brand]?
Run the questions on ChatGPT, Perplexity and Google AI Mode.
Do not rely on a single question or one screenshot. The wording of a question, the date of the search and the sources available can change the response.
Record the following information:
| What to record | Why it matters |
|---|---|
| Exact question | Makes the test repeatable |
| Platform | Shows where the issue appears |
| Date and time | AI answers can change |
| Exact statement | Identifies the claim that needs checking |
| Cited URLs | Helps locate the original information |
| Publication date | Shows whether the information is old |
| Accuracy | Separates false claims from genuine criticism |
| Business impact | Helps prioritise the response |
This becomes your baseline reputation report.
Trace the claim back to its original source
An AI summary is not necessarily the original publisher of the information.
The claim may have come from a newspaper report, a government notice, a customer complaint, a discussion forum, a company filing or another website.
Open every cited source. Check whether the source actually supports the statement made in the AI answer.
For example, a news report may say that a company investigated a customer complaint in 2024. The AI answer may describe the company as currently facing an unresolved investigation.
Those are two different statements.
The first is a historical event. The second suggests that the matter remains unresolved.
Before responding, check the original report, subsequent coverage and any official records available.
Do not assume that the AI has invented the entire story simply because its summary is misleading. Sometimes the underlying article is accurate but the summary misses a relevant update.
Correct the original information first
This is where reputation management often goes wrong.
A company publishes a new article saying that it has resolved a problem, but the original inaccurate report remains unchanged.
The new article may provide additional context, but it does not automatically correct the original source.
If the original publisher has made a factual error, contact its editorial or corrections team.
Provide:
- The exact sentence that is incorrect.
- The correct information.
- Documents or records supporting the correction.
- The date of the relevant event.
- A clear request for a correction or update.
Avoid sending a general complaint saying that the article is damaging your reputation. Explain precisely what is factually wrong.
If the information is accurate, do not demand that the publisher remove it simply because it is unfavourable. Instead, provide relevant developments and request that the article be updated where appropriate.
Build a reliable source of current information
Once the original facts are established, make sure accurate information is available on your own website.
A company website should not be limited to product pages and marketing messages. It should also provide reliable information about the organisation.
Useful pages include:
| Page | Information to include |
|---|---|
| About Us | Company history, leadership and business activities |
| Leadership | Current leadership and relevant professional details |
| Press and Media | Official announcements and media statements |
| Company Updates | Important developments and changes |
| Product Information | Current specifications, policies and product details |
| Customer Support | Complaint channels and resolution procedures |
| Investor Relations | Verified financial disclosures and official filings, where applicable |
If a company has addressed a serious issue, explain what happened, what action was taken and what evidence supports the resolution.
Do not write a vague statement such as “We are committed to customer satisfaction.”
Explain the actual change.
For example, if a company changed its refund policy after customer complaints, publish the revised policy, its effective date and the customer support process.
A specific, verifiable update is more useful than another promotional statement.
Understand what each platform allows you to report
There is no single reputation removal form that works across all AI platforms.
The reporting process depends on the platform and the type of problem.
Google allows users to provide feedback on AI Overviews when an answer is inaccurate or problematic.
For Knowledge Panels, verified representatives may submit feedback about information concerning the person or organisation they represent.
Google explains that Knowledge Panel information is generated from public sources. Its guidance also says that descriptions generally need to be corrected at their source. Submitting feedback does not guarantee a particular change.
References:
ChatGPT
ChatGPT can make incorrect or misleading statements. OpenAI advises users to verify important claims and check the sources provided in search responses.
Where content violates OpenAI’s policies or applicable laws, users can submit a report through its reporting channels. OpenAI says reported domains and content may be reviewed, and filters or other measures may be applied.
This is not a general process for removing every negative statement about a company.
References:
Perplexity and other AI search platforms
Check the relevant platform’s own reporting and feedback options. Keep the exact answer, source URL, date and supporting evidence ready.
Do not assume that a complaint submitted to one platform will change what another platform displays.
Measure whether the situation is improving
A company should not judge reputation recovery by counting how many positive articles it has published.
The more useful question is whether AI platforms are presenting a more accurate picture of the company.
Create a monthly measurement sheet.
| Metric | What to measure |
|---|---|
| Negative claim frequency | How often the specific inaccurate claim appears |
| Accuracy rate | Percentage of tested answers that are factually correct |
| Official source inclusion | How often relevant official information is cited |
| Context coverage | Whether important developments are included |
| Brand description accuracy | Whether the company is described correctly |
| Source freshness | Whether answers rely on current information |
| Resolution status | Whether the original publisher or platform has responded |
A simple accuracy score can be calculated as:
Accuracy rate = Number of factually accurate tested answers ÷ Total tested answers × 100
For example, if 16 out of 20 tested answers are accurate, the accuracy rate is 80%.
This is an internal monitoring measure, not an official AI platform metric. Keep the test questions consistent each month so that comparisons are meaningful.
A practical 30-day reputation recovery plan
| Period | Activity | Expected output |
|---|---|---|
| Days 1 to 5 | Test AI platforms and identify problematic claims | Baseline report |
| Days 6 to 10 | Trace claims to original sources and verify facts | Evidence register |
| Days 11 to 15 | Contact publishers about demonstrable errors | Correction requests |
| Days 16 to 22 | Update official company information and publish relevant evidence | Current information hub |
| Days 23 to 30 | Repeat AI platform tests and compare responses | First progress report |
Thirty days is a working review period, not a promise that AI answers will change within that time. External publishers and AI platforms control their own updates and response processes.
What should a brand do first?
If your brand is appearing negatively in AI search summaries, do not begin by trying to manipulate the answer.
Start with the evidence.
- Identify the exact claim.
- Check whether it is true, false or incomplete.
- Correct the original source where necessary.
- Publish documented updates on official channels.
- Use the reporting options available on the relevant platforms.
- Measure the answers again using the same questions.
The goal is not to manufacture a positive reputation. It is to make sure that people and AI systems can find accurate, current and complete information about your organisation.
A reputation cannot be repaired by content alone. The underlying issue has to be addressed first.
Sources
This article explains practical reputation management steps using publicly available platform guidance. It is not a report of an independent experiment conducted by this publication. Platform processes can change, so check the relevant official documentation before submitting a report.