Key Takeaways
- Four Google surfaces can now assert a fact about your business, and only two carry an accuracy warning. The AI Overviews help page says "AI responses may include mistakes" and that overviews "can and will make mistakes". The Ask Maps launch post carries no such caveat.
- Google documents no product-side removal for an AI answer, and the only route that names AI sits in the legal flow. Its legal troubleshooter lists AI Overviews and AI Mode as Search surfaces, and asks: "Does this request relate to content generated by AI within a Google product?"
- Thumbs-down is explicitly not a legal complaint. Google's wording: reporting through a content or product policy path "does not substitute for reporting it through a legal path and does not serve as legal notice", and each report must be filed separately.
- The two paths produce different maps. A legal win is normally national, applying "only in the country/region where it is deemed to be illegal". A policy breach is handled "globally".
- One court has already treated an AI Overview as Google's own statement. Munich Regional Court I granted a preliminary injunction on 28 May 2026, with penalties of up to EUR 250,000 per breach. The US case was sent back to state court on 9 January 2026 with nothing decided on the merits.
- Four Google surfaces can now make a claim about your business
- An AI claim is not a review, and that changes what you can ask for
- Capture it before it regenerates
- The three routes Google documents, and only one is legal notice
- The AI branch inside Google's legal removal flow
- Country-only or global: where the remedy lands
- What the courts have actually decided
- Q and A is gone, and the Ask button answers from your reviews
- Fix the input: where a wrong AI Overview about your business comes from
- The sixty-minute triage
- Frequently asked questions
A customer forwards you a screenshot: Google's AI has told them your business is under investigation, or closed, or known for something you have never been accused of, in a confident paragraph above the search results. This is worse than a bad review in one specific way, which is that there is nobody to answer. A review has an author and a flag button; a Google AI Overview wrong about your business has neither, and the sentence is gone from your screen the moment the model regenerates it. We read Google's AI Overviews help page, its legal reporting troubleshooter and its Ask Maps announcement line by line, pulled the court orders in the two cases that have actually tested this, and scored the five pages ranking for this query. Not one of them mentions the one place Google names AI output as reportable.
Four Google surfaces can now make a claim about your business
Almost every guide to this problem treats it as one feature. It is four, they have different inputs, and only two of them publish any warning that the answer might be wrong. That matters because your argument to Google, and later to a lawyer, rests partly on what Google itself said the thing was.
The two Search surfaces are documented and carry caveats. The two Maps surfaces are newer, draw on your reviews rather than on the open web, and have no published correction route whatsoever. Google's Ask Maps announcement of 12 March 2026 is the primary source for the third row, and the absence of any accuracy line in it is not an oversight on our part.
| Surface | What Google says it draws on | Accuracy wording Google publishes | Correction route Google names |
|---|---|---|---|
| AI Overviews in Search | Web pages it cites beneath the answer | "AI responses may include mistakes"; overviews "can and will make mistakes" | Feedback, plus a named option in the legal troubleshooter |
| AI Mode in Search | Web results, across a conversation | The same generative AI limitations page | A named option in the legal troubleshooter |
| Ask Maps (from 12 March 2026) | "Information from over 300 million places", community reviews, "insider tips from real people", personalised to the user | None in the launch announcement | None published |
| The Ask button on your profile, replacing Q and A | "Your answers and relevant reviews" | None published | None published |
Work out which row you are in before you do anything else, because rows one and two have a documented escalation and rows three and four do not. The remedy in rows three and four is upstream, in the reviews the answer is built from.
An AI claim is not a review, and that changes what you can ask for
A review is a third party's opinion in a container Google hosts, and the whole machinery of review disputes rests on that: you point at a published policy line, Google judges the review against it, and it stays or goes. A generated answer breaks all three parts. No author has violated a policy, there is no fixed text to judge, and there is nothing to delete.
That is why the reporting surfaces feel so thin. The thumbs-down form was built to improve a model, not to adjudicate a dispute, and Google says exactly that: feedback "helps us improve AI Overviews for everyone". There is no case number, no decision and no appeal, because nothing in the flow is deciding anything about you.
Two things follow, and they are the spine of this guide. First, your leverage is either the input (the review, the page, the directory entry the answer was built from) or the law (the claim is unlawful and Google is on notice of it). Second, those are different filings with different evidence, and a strong version of one does not help the other. If you have never separated the two, our guide to documenting evidence for a Google dispute is the right habit to bring here.
Capture it before it regenerates
Generated answers are not stable. The same query an hour later can produce different wording, a different set of citations, or no overview at all, and a complaint that cannot show the words is not a complaint. Do the capture first, every time, before you start fixing anything.
The exact query you typed, with the country and language you typed it in. A full-page screenshot showing the answer and the Google URL bar. The answer transcribed as text, so it can be quoted in a filing. Every source Google listed under it, as URLs. And the timestamp. Then repeat the query logged out, or on a second device, and note whether the claim survives: Ask Maps answers are "personalized based on things like places you've searched for or saved in Maps", and a claim only you can see is the weaker fact.
Keep the citation list in particular. It is the difference between a complaint about Google and a complaint about a page Google read, and those go to different places.
The three routes Google documents, and only one is legal notice
Here is the finding that reorders everything, and it is a sentence on Google's own reporting page rather than anything subtle. Google runs two parallel complaint systems, a policy one and a legal one, and it states plainly that the first is not the second: "You may report the same content through both legal and content/product policy reporting paths, but you must file each report separately. Note that reporting content through a content/product policy path does not substitute for reporting it through a legal path and does not serve as legal notice."
Read that against the standard advice, which is to click thumbs-down and wait. Every page ranking for this query leads with the feedback form. None of them says that the form you were sent to is, by Google's own description, not notice of anything.
| Route | What you are actually asking for | What Google says happens | Legal notice? | Scope of any remedy |
|---|---|---|---|---|
| Feedback: thumbs-down, then "Report a problem" | That this answer be improved | Feedback "helps us improve AI Overviews for everyone". No response time or outcome is published | No | None promised |
| Content or product policy report, through the product's own path | Enforcement of Google's own policies against the content | Where content breaches the policies or Terms of Service, Google "typically remove[s] or restrict[s] access globally" | No, expressly | Global |
| Legal removal request, through the Report Content On Google troubleshooter | Removal or restriction because the content is unlawful where you are | Google reviews it against local law and its own policies | Yes | Usually one country or region |
File the feedback anyway, because it costs a minute and it is the only lever on a merely wrong claim. Just do not count it as having told Google anything in a sense that matters later.
The AI branch inside Google's legal removal flow
The common claim, repeated on four of the five pages ranking for this query, is that Google offers no way to report an AI answer and no appeal. That was true of the help pages. It is not true of the Report Content On Google troubleshooter, which has been quietly rebuilt around this exact case.
Choose Google Search and the form offers four surfaces as radio buttons: Google Images, AI Overviews, AI Mode, and Other Search Features. Then, on the Search branch and on others, it asks a question in so many words: "Does this request relate to content generated by AI within a Google product?" The affirmative answer is pre-written for you, reading "Yes, such as AI Overviews or AI Mode".
On the reasons screen, four options appear: Intellectual Property, Court order ("Report content that a court decision has determined is unlawful"), Defamation ("Report content that defames you or your business/organization (typically a false statement damaging reputation)"), and Other. The defamation wording names businesses and organisations explicitly, so a company does not have to squeeze itself into a route built for individuals.
A route existing is not a remedy. You are asserting that a specific sentence is unlawful where you live, which is a legal claim with consequences for getting it wrong, and Google is free to disagree. What the route does give you is a filing that is on the record, dated, and aimed at the right product, which is precisely what a thumbs-down is not. The same distinction between a policy argument and a legal one runs through review defamation and where the legal line sits.
Country-only or global: where the remedy lands
This is the detail that decides which route to spend your effort on, and no competing guide mentions it. Google's reporting page sets out two different geographies for two different wins. On the legal side: "Typically, we remove or restrict access to the content only in the country/region where it is deemed to be illegal." On the policy side: "When content is found to violate Google's content or product policies or Terms of Service, however, we typically remove or restrict access globally."
So a successful defamation filing in Ireland can leave the same claim visible to a customer searching from Texas, while a successful policy complaint about the review underneath it removes that input everywhere. For a business whose customers are all local, national is enough. For anyone selling across borders, the policy route against the source is the stronger play even though it feels like the weaker one.
Nothing stops you doing both, and Google's wording anticipates it, as long as you file them separately.
What the courts have actually decided
Two cases are worth knowing, because the arguments in them are the arguments you will meet in any serious correspondence with Google. The honest summary is that no court anywhere has finally decided whether Google is liable for what its AI says about a business, but one has come close enough to order it to stop.
| Case | Court and date | What the AI said | Where it stands | What it establishes |
|---|---|---|---|---|
| LTL LED, LLC (Wolf River Electric) v. Google LLC | US District Court, District of Minnesota; remand order 9 January 2026 | That the solar installer was facing a lawsuit from the Minnesota Attorney General, which the complaint says none of the cited sources reported | Sent back to Minnesota state court as Google removed it too late; nothing decided on the merits | A US business can get a defamation claim over an AI Overview in front of a court. Damages claimed were put at roughly USD 110m to 210m |
| Case 26 O 869/26 | Landgericht Munchen I, 26th civil chamber; hearing 23 April 2026, injunction 28 May 2026 | That a publisher "was known for disreputable business practices" and was "frequently perceived as operating a fraud scheme" | Preliminary injunction, appealable and not confirmed by any higher court | A court has treated an AI Overview as Google's own content, rejected the hosting defence, and attached penalties of up to EUR 250,000 per breach |
The Munich court's reasoning is the part to borrow. It held that a feature which synthesises information into an independent answer, rather than listing links, is making its own statement, that the claims appeared in none of the sources cited, and that an "AI-generated" label does not move the duty to check onto the reader. It also noted that Google had been notified by letter and through the reporting form, and the claim was still appearing days later. In the Minnesota case, Google's defence is that it is a platform rather than the publisher. Both halves of that argument will be quoted back at you, which is the practical reason to keep a dated record of having put Google on notice.
Q and A is gone, and the Ask button answers from your reviews
If you have looked for the Questions and answers section on your profile lately and failed to find it, that is not a bug. Google's My Business Q and A API was discontinued on 3 November 2025, and on 3 December 2025 Google confirmed it was withdrawing the feature from Maps, replacing the threads with an AI-powered Ask button.
What matters for this guide is Google's description of where the replacement gets its answers. Instead of browsing existing questions, a customer gets "an updated, instant answer based on your answers and relevant reviews". Read that slowly if you have ever had a fake review you could not get removed. The claim inside it is now an input to a sentence Google will say to your next customer, in Google's voice, with no star rating to discount it and no author for the reader to judge.
Google has published nothing about previewing those answers, correcting an inaccurate one, or how often the underlying knowledge updates. The wider shift to AI in Google's review moderation is the same trade in a different direction: more automation, less adjudication, and fewer places to argue.
Fix the input: where a wrong AI Overview about your business comes from
Because there is no text to edit, the work is upstream. Sort every source on your captured citation list into one of these five rows. Be honest about the last one: sometimes no single source says the claim, and then only feedback or law remains.
| Input the answer was built from | Who controls it | How you change it | Realistic lag |
|---|---|---|---|
| Your own website | You | State the correct fact plainly on the page that ranks, remove the contradicting wording, then request a recrawl | Days to weeks |
| Business Profile fields | You, subject to Google's edits | Correct hours, categories, service area and attributes; check for a duplicate or merged listing | Hours to days |
| A review on your profile | The reviewer, or Google where it breaches a policy | Flag it against a named policy line with evidence; this is the only route that removes the claim at source | Days to weeks |
| A third-party article or directory entry | The publisher | Ask for a correction in writing; escalate legally only if the content is itself unlawful | Weeks |
| Nothing identifiable: the claim is synthesised | Feedback, then the legal route if the claim is unlawful where you trade | Undocumented |
Row three is where most of these cases actually live, and it is the row the SEO-flavoured advice skips. If the AI is repeating an accusation that a single review invented, the review is the defect, and the test for whether it is removable is the ordinary one in what counts as a fake Google review.
The sixty-minute triage
Do these in order, once, before anyone drafts a letter.
- Capture it. Query, locale, screenshot, transcription, citation list, timestamp. Repeat logged out.
- Name the surface. AI Overview, AI Mode, Ask Maps or the profile Ask button. The route differs by row.
- Classify the claim. Merely wrong, or a false statement of fact that damages you? Only the second has a legal route.
- Trace every citation. If a source says it, your fight is with that source, not with Google.
- Check for the boring explanations. A duplicate listing, a merged profile, a similarly named competitor, your own outdated page.
- File the feedback. Thumbs-down, "Report a problem", quote the false sentence, give the correct fact, name the page that proves it.
- Fix or dispute the input. Your page, your profile fields, or the review, with evidence against a named policy line.
- Only then consider the legal filing, through the troubleshooter's AI branch, with advice if the sums are serious.
Step four is the one that saves the most wasted effort, and step seven is the one that actually changes what the AI says next month.
The AI repeats the review. Remove the review.
Flaggd reads Google's policies against the actual content and files evidenced, single-ground disputes. When a generated answer is restating a review that breaches a policy line, the review is the only part of the chain anyone can have taken down, and a policy removal applies everywhere rather than in one country.
Flaggd is our own service. We dispute content that breaches Google policy; we do not remove or suppress content that does not, nobody can guarantee a removal, and nothing here is legal advice.
Frequently asked questions
Can I get a wrong Google AI Overview about my business removed?
There is no setting that deletes an AI Overview, because the text is generated on demand rather than stored. Google offers feedback through the thumbs-down control, and its legal reporting troubleshooter now lists AI Overviews and AI Mode as reportable Search surfaces, with defamation as a selectable reason. In practice most cases are solved upstream, by correcting or removing the source the answer was built from.
Does the thumbs-down feedback form count as a legal complaint to Google?
No, and Google says so in writing: reporting through a content or product policy path "does not substitute for reporting it through a legal path and does not serve as legal notice", and each report must be filed separately. To put Google on notice in a sense that matters later, use the legal troubleshooter as well.
How do I report an AI Overview under Google's legal process?
Open the Report Content On Google troubleshooter, choose Google Search, then choose AI Overviews or AI Mode from the list of Search surfaces. It then asks whether the request relates to content generated by AI within a Google product, with a prepared answer reading "Yes, such as AI Overviews or AI Mode". On the reasons screen, Defamation covers content that "defames you or your business/organization".
If Google agrees to remove something, does it disappear worldwide?
It depends on the path. Google says that on a legal request it typically removes or restricts access "only in the country/region where it is deemed to be illegal", while content found to breach its own policies or Terms of Service is "typically" removed or restricted "globally". For a business with customers in several countries, the policy route against the underlying source reaches further.
Has any business won a case against Google over an AI Overview?
No case has been finally decided. Munich Regional Court I granted a preliminary injunction on 28 May 2026 in case 26 O 869/26, holding that an AI Overview was Google's own content and attaching penalties of up to EUR 250,000 per breach. It is preliminary and appealable. The Minnesota claim was remanded to state court on 9 January 2026 on a procedural point.
Why does Google's AI repeat an accusation from a review I believe is fake?
Because reviews are an input. Google says the AI answer replacing Business Profile questions gives customers "an updated, instant answer based on your answers and relevant reviews", and the Ask Maps launch describes drawing on community reviews and "insider tips from real people". A claim invented in one review can be restated in Google's own voice, with no star rating and no author beside it, so removing the review is the only fix that stops the restatement.
What happened to the Questions and answers section on my Business Profile?
Google discontinued the My Business Q and A API on 3 November 2025 and confirmed on 3 December 2025 that it was withdrawing the feature from Maps, replacing question threads with an AI-powered Ask button. Previously published answers continue to inform what Google shows. Google has not published how to preview those answers, correct an inaccurate one, or how often the knowledge refreshes.
How long does it take for a wrong AI answer about my business to change?
Google publishes no timeline and no outcome for feedback, so treat any estimate as a guess. What is observable is the lag on the inputs: profile field corrections tend to show within hours or days, a recrawled page or a review removal within days or weeks. Re-capture the query weekly rather than assuming one corrected result means it is fixed.
The hard part of this problem is that the thing saying it has no author to be reasonable with. The workable part is that it is built out of material that does have authors, some of it breaching rules Google already enforces. Treat the sentence as a symptom, find what it is reading, and you are back on ground where a documented filing works.