AI-generated answers can get a brand’s name, product, ownership, pricing, policies, or history wrong. The practical response is to document the exact answer, verify the fact against a reliable source, correct inaccurate source material where you control it, send platform feedback through an appropriate route, and recheck a consistent prompt sample later. A report is a request for review—not a guarantee that an answer will change.
This guide covers consumer-facing AI answers and public brand facts. It does not replace legal advice or a platform’s current reporting instructions. It also does not assume that every mistake comes from a model’s training data: a response may reflect a live search result, a third-party page, an ambiguous prompt, or a one-off generation.
What counts as an AI-generated brand error?
An AI brand error is a checkable, materially inaccurate statement about an organization, product, person, or policy. Examples include an obsolete price, a feature attributed to the wrong product, a former executive presented as current, a false company relationship, or a citation that does not support the nearby claim.
Separate factual errors from subjective summaries. “The answer does not describe us the way we prefer” is not, by itself, evidence of a false statement. Write down the exact sentence and the verifiable fact that contradicts it before deciding what action to take.
What evidence should you save first?
Capture the answer before you try to correct anything. Save the platform and surface, the prompt, date and time, locale and language, search or answer mode, the full relevant response, and every displayed citation URL. A screenshot can help preserve the interface, but keep the text and source URLs in your record too.
Re-run the same prompt in a fresh session if practical. Note whether the issue repeats, changes, or disappears, and keep all runs—including answers with no brand mention. One output shows what happened once; it does not establish a stable platform behavior or a broad visibility rate.
How do you trace the incorrect fact to its source?
Check each cited page and the brand’s own current public pages. Ask whether the claim appears in a source, whether that source is outdated, whether two entities have been confused, or whether the answer added a detail that none of the sources support. A visible citation is evidence of a displayed source, not proof that the source caused every part of the answer.
Classify the issue before acting:
- First-party page is wrong or stale: correct the page, date, pricing, policy, product detail, or organization fact at its source.
- Third-party source is wrong: contact its publisher with the exact passage and a credible, public correction source. Keep the request factual and specific.
- Entity confusion: clarify the distinction on pages you control using consistent names, descriptions, and links to genuine official profiles. Do not create profiles or structured data that imply relationships that are not real.
- Unsupported answer claim: preserve the answer and citations. If no source supports the statement, describe that precisely in product feedback.
- Ambiguous or subjective wording: rewrite the prompt as a neutral, testable question and check whether the answer still makes a factual assertion.
How can you report an inaccurate answer?
Use the platform’s feedback path that matches the issue and include only the evidence needed to identify it. Interfaces change, so follow the controls shown in the product when you submit.
Google AI Overviews
Google’s Search Help says to use the thumbs-down control on an AI Overview, choose Report a problem, select the category, add details, and submit. Google says the feedback includes the most recent query and search results. Include a concise correction and a reliable supporting URL; avoid adding customer or confidential information. Feedback helps Google evaluate the feature, but Google does not promise that a particular overview will be corrected or removed.
Perplexity
Perplexity’s Help Center says users can use the flag icon below an answer or contact support, and recommends including the query URL, a description of the error, and the expected result. It lists misinformation and outdated information among example issues. Confirm the current in-product control before sending because help pages and interface labels can change.
ChatGPT
OpenAI documents thumbs-down feedback on an individual ChatGPT response. Its formal content-reporting route is for material that may violate OpenAI’s Terms of Use or applicable law; the reporting page says reported domains or content may be reviewed and mitigations may be applied to help prevent reliance on unreliable sources. Those are review possibilities, not a fact-correction service or a promised outcome. Do not characterize an ordinary disagreement as a policy or legal violation to try to force a correction.
Other AI products
Look for the product’s own response feedback, support, or content-reporting instructions. Do not assume one platform’s flag, appeal, or crawler process applies to another provider. If no factual-feedback route is documented, record the issue and correct the source material you control rather than submitting a misleading report.
What should a useful correction report contain?
Keep the report short and reproducible. Include:
- The exact prompt and answer URL, if the product provides one.
- The specific sentence that is incorrect and a one-sentence explanation.
- A dated, authoritative source showing the accurate public fact.
- The answer’s citation URL, if it points to an incorrect or outdated source.
- The date, product, mode, locale, and whether the issue repeated.
Do not flood the service with repeated reports or ask other people to manufacture agreement. If the error concerns personal data, safety, impersonation, copyright, or a legal right, use the platform’s dedicated route for that category and seek qualified advice where needed.
How do you check whether the answer changed?
After correcting a source or submitting feedback, re-run the same prompt set under comparable conditions and record the date. Track the original wording, citation, mention, and result separately. Answers can vary, search indexes can update on different schedules, and a feedback submission does not reveal what action—if any—the provider took.
Use the same sampling rules you use for AI visibility measurement and keep mentions, citations, recommendation context, and referrals distinct as described in AI Visibility Metrics Explained. Report the number of runs and the time period; do not treat a single changed answer as proof of a permanent correction.
A careful response checklist
- Preserve the exact answer, prompt, citations, platform, and date.
- Repeat the prompt to check whether the error is reproducible.
- Verify the disputed statement against an authoritative public source.
- Correct the source you control and ask third-party publishers for specific factual updates.
- Use the platform’s documented feedback route; do not imply that feedback guarantees a change.
- Retest the same prompt set later and report the sample and limits.
For the upstream mechanisms and evidence limits behind visible citations, see How AI Search Engines Choose Sources: What the Evidence Shows. For the broader site approach, start with How to Improve AI Visibility.
Frequently asked questions
Can a brand force an AI answer to be corrected?
There is no documented universal process that guarantees an AI-generated answer will be changed. You can correct public sources you control, submit accurate feedback through a platform’s route, and monitor later outputs, but the provider controls its systems and response.
Should I report every answer that does not recommend my brand?
No. A missing recommendation or an unfavorable opinion is not automatically a factual error. Report a specific, checkable inaccuracy through a suitable channel, and keep commercial preferences separate from factual corrections.
Does fixing my website immediately change AI answers?
Not necessarily. A product may use a live search index, other sources, or model knowledge, and providers do not promise a common refresh time. Correct the underlying source, then measure comparable outputs over time.
What if a third-party page is the source of the mistake?
Contact the publisher with the exact passage and a reliable source for the correction. Save the request and any response, then monitor the page and the AI answer separately. Do not claim a publisher has corrected something until you verify the live page.
Sources and update note
- Google Search Help: AI Overviews and feedback
- Perplexity Help Center: Need support?
- OpenAI Help Center: Reporting Content in ChatGPT and OpenAI Platforms
- OpenAI Help Center: Does ChatGPT tell the truth?
Platform controls and policies can change. This guide was checked against the linked public help pages on September 28, 2026. It describes feedback and correction steps, not guaranteed model, ranking, or citation outcomes.