Key Takeaways
- You cannot look up your own removed review, and one field settles it. The only content identifier a statement of reasons accepts is an EAN-13 product barcode. Nothing in the record points at a review, a profile, a place or an account.
- Google Maps files two structurally different kinds of decision, and they have different appeal routes behind them. All 1,000 of the newest illegal-content rows name Defamation and arrive as user notices. All 1,000 of the newest text-content rows name a Google policy and are Google's own initiative.
- 995 of the newest 1,000 rows filed under "Google Maps" were adverts, not Maps content. Set Content type to Text and the advertising boilerplate disappears completely, which is the single filter that makes the platform row readable.
- The defamation removals are almost entirely German. On one day we counted 969 rows scoped to Germany alone against zero for Ireland, Italy and France, measured from the database rather than from anyone's summary of it.
- The export stops at 1,000 rows, and the newest 1,000 covered five minutes. Without a date filter typed as DD-MM-YYYY you are reading the last few minutes of Google's submissions and nothing else.
- What the database is, and what Article 17 actually promises
- Can you find your own review in it? No, and one field settles it
- Google Maps files two completely different kinds of decision
- Why 995 of the newest 1,000 rows are adverts
- The one filter that isolates Maps content decisions
- The policy grounds Google actually files under
- The defamation rows are almost entirely German
- Why the search cannot reach last week
- What we would actually do with it in the first week
- Frequently asked questions
A review disappeared from your profile three weeks ago, Google sent you nothing, and the support agent you reached could only read you the policy list. Meanwhile the European Commission has been publishing Google's own filed reason for every Maps moderation decision since September 2023, which means a version of the explanation you were refused sits in a public database that anyone can search. The promise of this guide is narrower than that sounds, and we are going to be straight about it: we spent this morning querying the DSA transparency database for Google Maps, pulling three full exports and five one-day country slices, and your specific review is not findable in any of them. What is findable is better than nothing and worse than the headline suggests.
What the database is, and what Article 17 actually promises
Article 17 of the Digital Services Act is the part that matters to a business owner, and it is not really about databases at all. It says that when a hosting service restricts content, it owes the affected party a statement of reasons: the restriction applied, its territorial scope and duration, the facts and circumstances relied on, whether automated means were used, and the legal or contractual ground. Article 24(5) then requires online platforms to send those statements to the Commission, which publishes them. The Commission's own description is that it "collects and makes publicly available statements of reasons", and it has done so since 25 September 2023.
So there are two obligations, running in opposite directions, and conflating them is where most advice on this goes wrong. The statement of reasons goes to the person whose content was restricted, which for a removed review means the reviewer, not you. The copy that goes to the Commission is deliberately stripped: Article 24(5) requires platforms not to include personal data, and the Commission's guidance tells them to be "particularly careful with free text fields". The public database is therefore the same decision with every identifying thread cut.
Worth holding onto: Article 17 does not cover orders from authorities, which are Article 9, and it does not cover the notice someone files to start the process, which is Article 16. The database records what the platform decided, not what anyone asked for. If you reported a review and want to know what happened to your report, this is the wrong instrument, and the status tool on your own profile is the right one.
One more design decision shapes everything below. The Commission's guidance on the facts-and-circumstances field says platforms "are encouraged to use standard language for similar cases". Google took that invitation seriously.
Can you find your own review in it? No, and one field settles it
The schema has one attribute for identifying the content a decision was taken against. It is called content_id, and the Commission's explanation of the statement attributes states that "the only accepted key-value combination is a product identifier, namely the EAN-13 code". An EAN-13 is a retail barcode. There is no field for a URL, a place ID, a review ID, a reviewer name or a business name, and across all 3,000 rows we exported, content_id was empty in every single one.
That is not an oversight, it is the point. The record keeps the shape of the decision and discards the subject. You get the ground, the date, the territory, the automation flag and a category. You do not get the thing. So the honest framing for the rest of this guide is not "how to find your review" but how to read the population your review fell into, which turns out to tell you something your support agent will not.
There is one identifier in the published row, a field Google fills called platform_uid, carrying values like 564968858692. It is Google's internal reference for the statement, not for the review, and nothing you can see from your side of the profile will ever match it. Treat it as a citation handle for a row you have already found, not a way in.
Google Maps files two completely different kinds of decision
Google Maps appears in the platform list as its own entity, separate from Google Play and Google Shopping. Export its rows and sort them by decision ground and the population splits cleanly in two, with almost no overlap in any other field. This is the most useful thing in the database for an owner, because which half your removal landed in decides what recourse you have.
| Field as filed | Illegal-content rows (newest 1,000) | Terms-and-conditions rows (newest 1,000 text) |
|---|---|---|
| Legal or policy ground named | Defamation, 1,000 of 1,000 | 23 distinct Google Maps policy names |
| Who started it | Article 16 notice, 1,000 of 1,000 | Google's own initiative, 1,000 of 1,000 |
| Automated decision | Not automated 589, fully automated 411 | Fully automated 994, partially 6 |
| Territorial scope | Germany alone 983, EEA-wide 15, Austria 1, Ireland 1 | EEA-wide, 1,000 of 1,000 |
| Content type as filed | Other, 1,000 of 1,000 | Text, 1,000 of 1,000 |
| Statement category | Illegal or harmful speech, 1,000 of 1,000 | Scams and fraud 707, other T&C 190, speech 57, data protection 41, cyber violence 5 |
| Minutes of filing the 1,000 covered | 6 hours 10 minutes | 5 minutes 23 seconds |
Read the automation row again, because it is the one with consequences. A defamation removal is a human decision more often than not, and a human decision came from somebody's notice: there is a complainant, and there is a legal theory you can answer. A terms-and-conditions removal is machine work at a rate of 994 in 1,000, triggered by nothing but Google's own detection. Those are not the same problem, and the line between a policy breach and an unlawful statement is where the two diverge.
Why 995 of the newest 1,000 rows are adverts
Here is the trap that makes a first look at this database useless, and nothing we could find online warns about it. Export the newest 1,000 Google Maps rows with no filters at all and 995 of them are advertising decisions. We know because the facts-and-circumstances text gives it away: 995 of the 1,000 carried an identical paragraph describing how Google weighs "the content of the creative (e.g. ad text, keywords, and any images and video) as well as the associated ad destination".
The grounds confirm it. The top entries in that unfiltered export were Ad format requirements (357), Destination requirements (328), Abusing the ad network (158), Technical requirements (45), Healthcare and medicines (40) and Alcohol (39). Those are Google Ads policies. They have nothing to do with a review on your profile, and a casual reader who filters to Platform equals Google Maps, sees "Scams and fraud" as the second-biggest category and concludes something about review fraud has drawn exactly the wrong lesson from the data.
The five survivors are instructive. Of the 1,000 unfiltered rows, five carried the generic content paragraph instead of the advertising one. Their grounds were Vandalism, Fake engagement and Sexually explicit content, and their explanation fields held the notice text in the contributor's own language, French and Ukrainian among them. One of the French ones reads, in translation, that automated systems removed the contribution because it appears not to comply with the Maps content rules. That is the user-facing message, stored verbatim.
So the platform label is doing less work than it appears to. "Google Maps" in this database is the legal entity's filing bucket, and advertising volume swamps everything else in it.
The one filter that isolates Maps content decisions
Set Content type to Text and the advertising rows vanish completely. We re-ran the same export with that single change and all 1,000 rows came back carrying the generic content paragraph, with zero occurrences of the advertising boilerplate. The grounds changed from ad-policy names to Maps content-policy names in one step. That is the whole trick, and it is worth more than every other filter on the page combined.
| Filter | Set it to | Why, from what we measured |
|---|---|---|
| Platform | Google Maps | Google Play and Google Shopping are separate entries. Reviews on business profiles are filed under Maps. |
| Content type | Text | Removes 995 advertising rows per 1,000. Non-negotiable. |
| Decision ground | Leave blank, then run it twice | The two populations never mix. Illegal-content rows are filed as Other, not Text, so Text alone hides them. |
| Created starting / ending | DD-MM-YYYY, never ISO | An ISO date is accepted and ignored. You get the newest rows as though you had set nothing. |
| Territorial scope | Your own country | It matches rows whose scope includes that country, so EEA-wide rows still appear. It narrows, it does not exclude. |
One caveat we have to put next to that recipe, because it limits the conclusion you can draw. Text content on Google Maps is not only reviews. Place edits, answers and other contributions are filed the same way, and the Vandalism row above is plainly an edit rather than a review: its notice text refers to a contribution that could not be published as a modification. There is no field that separates a review from a suggested edit, so treat the text population as Maps user contributions generally, with reviews as the bulk of it rather than the whole of it.
The policy grounds Google actually files under
This is the part that earns the trip. Filtered to text content, the newest 1,000 Maps rows name 23 distinct grounds, and they are recognisably Google's own review policy headings. The distribution is lopsided in a way that should change how you read your own missing review.
| Ground, as Google files it | Rows per 1,000 | What it covers |
|---|---|---|
| Fake engagement | 613 | Bought, incentivised or coordinated activity |
| Posting Restrictions (plus 18 under translated names) | 115 | An account or place restricted from receiving posts |
| Misrepresentation | 47 | Impersonation and false identity or experience |
| Personal Information (two spellings combined) | 41 | Staff names, phone numbers, medical detail |
| Unclear content | 33 | Gibberish and unintelligible submissions |
| Offensive Content (plus 1 translated) | 31 | Shocking or gratuitously offensive material |
| Obscenity & Profanity | 25 | Vulgar language |
| Prohibited and Restricted Content | 21 | Regulated goods, including the incentives rule |
| Sexually explicit content (plus 1 translated) | 13 | Nudity and descriptions of sexual acts |
| Vandalism | 13 | Malicious edits to a place entry |
| Conflict of interest owner | 13 | Self-reviews and reviews of a rival |
| Harassment / Adult-themed / Off-Topic / Repetitive / Hate speech | 17 combined | The long tail, each in single figures |
Two things leap out. The first is that Fake engagement is 61% of all text-content removals on its own, a very different picture from the one most reputation advice paints, where off-topic and profanity do the heavy lifting. The second matters for anyone searching this database: the ground names are not normalised. "Personal Information" and "Personal information" are separate values in the same export, and Posting Restrictions also appears as French, Italian, Swedish and Dutch strings, so a free-text search for a ground name will silently miss rows.
Note also what is almost absent. Off-Topic appeared three times in a thousand, which does not mean off-topic reviews are rare; it means they are rarely the ground Google files under when it removes something itself. That gap between the policy list you report against and the grounds Google records is the most quietly useful finding here.
The defamation rows are almost entirely German
Switch the decision ground to illegal content and the database stops looking like a content-policy log and starts looking like a legal docket. In the newest 1,000 illegal-content rows for Google Maps, the legal ground named was Defamation in all 1,000. Not one other legal theory appeared: no privacy claim, no consumer law, no trade-mark complaint. And 983 of those 1,000 were scoped to Germany alone.
Secondary reporting has circulated a claim that almost all of Google's EU defamation removals are German, and we did not want to repeat a number we had not checked, so we measured it. We took a single day, 20 October 2026, filtered to illegal-content rows for Google Maps, and ran the territorial scope filter once per country.
| Territorial scope filter | Rows returned | Scoped to that country alone | EEA-wide rows in the result |
|---|---|---|---|
| Germany | 1,000 (export cap) | 969 | 31 |
| Austria | 88 | 4 | 84 |
| Ireland | 84 | 0 | 84 |
| Italy | 84 | 0 | 84 |
| France | 84 | 0 | 84 |
The 84 EEA-wide rows are the same 84 in every country's result, which is why Ireland, Italy and France return an identical count: they share a pool and contribute nothing of their own. Germany contributed at least 969 country-specific defamation removals in one day, and the only reason we cannot say how many more is that the export stops at a thousand. Austria managed four.
The practical reading for a business outside Germany is blunt. The legal route to a removal is barely used where you trade, which is both a warning and an opening: your competitors are not using it either, and the DSA appeal route against a refusal is a separate mechanism that nothing in this table measures.
Why the search cannot reach last week
The database page states its own limits plainly: results "cover the last 6 months of data submitted", data is "updated once every day", "only the first 10 000 results are paginated", and the first 1,000 can be downloaded as a CSV. Those numbers sound generous until you measure Google's filing rate against them.
Our newest 1,000 text-content rows were all submitted between 22:38:34 and 22:43:57 on one evening. Five minutes and twenty-three seconds of filings exhausted the entire export. The illegal-content thousand covered six hours and ten minutes. So if your review disappeared a fortnight ago, the default view will never show you anything from that day, and no amount of scrolling helps: the pagination ceiling of 10,000 runs out long before a fortnight does.
Three mechanical traps, all of which cost us time this morning:
- The date fields take DD-MM-YYYY and silently ignore anything else. We first typed 2026-10-01 in the ISO format the rest of the world uses. No error appeared. The result was the newest rows, identical to running no filter at all, which is the worst possible failure mode because it looks like an answer.
- A one-day filter still hits the cap on a busy ground. Germany plus illegal content plus a single date returned exactly 1,000. You have to add a narrower filter, not a narrower date, to get an enumerable set.
- Some queries return HTTP 500 and then work on a retry. One date that failed outright succeeded unchanged a few minutes later. A server error here is not evidence that no data exists for that day.
The six-month retention is the hard wall. A review removed last year is gone from this database whatever you filter, which makes it useless for the historic audit most owners actually want and reasonable for the question "what is happening to profiles like mine right now". That is also why a sudden drop in your review count is better diagnosed from your own records than from here.
What we would actually do with it in the first week
The database is not an evidence source for a dispute. It is a calibration tool, and used that way it is worth an hour. Here is the order we would work in.
- Fix your own date first. Write down when the review vanished, from your screenshots or notification emails: every query below depends on a date the database will not supply.
- Run the text-content query for your country and that week. Platform Google Maps, Content type Text, territorial scope your country, dates in DD-MM-YYYY. Read the grounds distribution, not the individual rows.
- Run it again with the ground set to illegal content. If your country returns only the shared EEA-wide rows, as Ireland, Italy and France did, then nobody is filing legal complaints about reviews where you trade and a defamation framing is unlikely to be what happened to you.
- Compare the grounds against the one you assumed. If you were about to report a rival's reviews as off-topic, Fake engagement outnumbering Off-Topic by roughly two hundred to one is a reason to re-read your evidence first.
- Then stop, and go and work on your own profile, which is the only surface with a documented status lookup and an appeal behind it.
Set expectations honestly, because this is where the guide parts company with the hopeful version. Nothing here will name your review, prove who complained, or give you a case to cite in an appeal. It tells you which of two machines took your content, and how often that machine handles things like yours. That is calibration, and calibration is what stops a business filing the wrong complaint for a quarter.
The ground you file under decides the outcome
Google's own filings show how narrow the real grounds are. Flaggd reads the reviews on your profile against Google's content policies and files the evidenced dispute under the ground that actually fits, rather than the one that feels strongest.
Flaggd is our own service and this page recommends it. We dispute content that breaches Google policy on profiles our clients control. We do not file statements of reasons, we cannot retrieve a database row for your review because no such row is identifiable, and nobody can guarantee a removal. Every figure above was queried from the Commission's database on 5 November 2026 and will have moved since; the six-month retention window means the exact rows we counted will expire from it.
Frequently asked questions
Can I find out why Google removed my review from the EU database?
Not for a specific review. The statement-of-reasons schema has one content identifier field, and the Commission's own documentation says the only accepted value is an EAN-13 product barcode, so no record points at a review, a profile or an account. Across the 3,000 Google Maps rows we exported, that field was empty in every one. What you can read is the population your removal fell into: the ground, the territory, whether a human or a machine decided, and whether a user notice started it.
What is the DSA transparency database?
It is a public database the European Commission has maintained since 25 September 2023. Article 17 of the Digital Services Act requires a hosting service to give a statement of reasons when it restricts content, and Article 24(5) requires online platforms to send those statements to the Commission, which publishes them in anonymised form. Anyone can search, read and download them without an account.
Can I search the DSA transparency database for my own business?
No. There is no business name, place identifier or URL in a published statement of reasons, and Article 24(5) requires platforms to leave personal data out of their submissions. The nearest you can get is to filter by platform, content type, decision ground, territorial scope and date, which narrows the set to decisions that resemble yours rather than to yours.
Is Google Maps in the DSA transparency database?
Yes, as its own platform entry, separate from Google Play and Google Shopping. Reviews and other contributions on business profiles are filed under Google Maps. Be aware that the entry also carries Google's advertising decisions, which outnumber Maps content decisions heavily in the default view.
Why are most Google Maps entries in the database about adverts?
Because the platform entry is a filing bucket for the entity rather than for the product. In the newest 1,000 unfiltered Google Maps rows we exported, 995 carried the advertising boilerplate describing how Google weighs ad text, keywords, images, video and the ad destination, with grounds like Ad format requirements and Abusing the ad network. Setting Content type to Text removed all of them and left only Maps content decisions.
How far back does the DSA transparency database go?
The search covers the last six months of submitted data, the set is updated once a day, only the first 10,000 results are paginated and only the first 1,000 can be exported. Those limits bite harder than they sound: the newest 1,000 Google Maps text-content rows we pulled covered five minutes and twenty-three seconds of filings, so without a date filter you are reading the last few minutes. A review removed more than six months ago is not in the database at all.
Why are almost all Google review defamation removals German?
We can confirm the pattern but not explain it from the data. In the newest 1,000 illegal-content rows for Google Maps the legal ground was Defamation in all 1,000, and 983 were scoped to Germany alone. Taking a single day and filtering country by country, Germany returned at least 969 country-scoped rows while Ireland, Italy and France each returned zero of their own and Austria returned four. Germany's law on business reviews and a well-developed legal practice around them are the usual explanation offered, but the database records decisions, not reasons for national differences.
Does a statement of reasons help me appeal a Google review decision?
Not directly, because you cannot identify the row for your own case and the published facts field is boilerplate. The Commission's guidance encourages platforms to use standard language for similar cases, and Google does: across 1,000 text-content rows the facts field held a single identical paragraph. Use the database to calibrate which ground is plausible, then appeal through the route that fits, which is Google's own Maps content reporting on a profile you own, the enforcement figures it publishes for context, or the DSA out-of-court mechanism after a refusal.