The things worth knowing first

Before you upload

Does anything change in my shop?

No. A scan reads a copy of your file and reports on it. A fix produces a separate file you add yourself in Merchant Center. Your CMS, your export, and your primary data source are never written to. The mechanism behind that is on how it works.

Do you need access to my Merchant Center account?

No. You give us a file and we give you a file back. There is no permission to grant and no code to install. The trade-off is that we can tell you what your data says, but not what Google finally decided about it, because that would need account access we do not ask for.

What file formats can you read?

csv, tsv, xlsx, xml and json, and zip or gz archives of any of those. Column headers in Dutch, French, German and English are recognised without configuration.

How large can my file be?

Up to 60 MB. If yours is larger, split it or get in touch. The public scan accepts up to 5,000 products. For a larger catalogue, contact us before uploading.

Does the scan use an AI model?

The free scan does not call a model at all. It is deterministic: a GS1 check digit is arithmetic, image dimensions are a fetch, and a robots rule is a file on your server. The same file produces the same report every time. A model is only involved in the paid rewriting of thin titles, and even then every proposed value is checked word by word against your own row.

Could I not just do this with a chatbot myself?

For the language parts, partly, and we use a model for those ourselves. The rest is not a language job. Whether a barcode passes its check digit is arithmetic a chatbot will happily wave through. Whether an image is 480 by 480, or answers at all, is only knowable by fetching every URL. Whether AdsBot-Google is blocked sits in a file on your server.

The deeper problem is that asking a model only fixes the faults you already knew to ask about, and leaves the ones you did not. The usual way to discover the rest is to upload and let Google refuse them, one slow round at a time, while your shop re-exports over your work in between. A scan runs the whole list in one pass and checks that the corrections do not undo each other.

Are you part of Google?

No, and nothing here is endorsed by them. What is behind it is years of Google Ads work across a lot of Shopping accounts and a lot of feeds with errors in them. This is the checklist that kept being rebuilt by hand.

After the scan

What if my feed turns out to be fine?

Then you pay nothing. The scan says so, there is nothing to download, and that is a good outcome rather than a wasted one.

What do I actually have to do in Merchant Center?

Put the corrections into a spreadsheet in your own Google account and add that sheet as a supplemental data source, attached to the right primary source. Google reads it every 24 hours from there. The step-by-step is on install.

What happens when my feed regenerates tonight?

Nothing is undone. Google lays the supplemental source over your primary source every time it processes it, matched on id, so your export can regenerate as often as it likes. That is the reason this is a one-time purchase and not a subscription.

Do you change my prices?

No value of yours is ever written into anything by hand. The scan reports format faults such as a comma decimal or a missing currency; the correction is applied as an overlay on the attributes we found a reason to change, matched on id, never by editing your file.

I add new products every week. Are those covered?

The per-product corrections cover the products that were in the file we read. Products you add later are not covered until you run a fresh scan that includes them.

The report is the other half of the answer. Most faults are not one-offs: an export tends to make the same mistake in the same field every time, and the report says which faults are systematic and where they come from, so you can fix them at the source and have new products arrive clean. Scanning again is free.

How soon will I see the difference?

Not immediately. Google refetches your sources on its own schedule, reprocesses the products, and lifts the disapprovals it agrees with. Your part is done the moment the data source is in place; the rest is Google's timing.

How long does a scan take?

Roughly a minute on a catalogue of a few thousand products. Most of that minute is spent fetching your images, one network request per product image, sixteen at a time, because that is the part a spreadsheet cannot tell you and it is worth waiting for.

Which column-header languages do you understand?

Dutch, French, German and English headers are recognised without any configuration, including the WooCommerce attribute-pair pattern. You do not have to rename your columns to Google's names first; the mapping works out what your headers mean.

About the findings

What do the severities mean?

They are our own judgement, read off the check that produced them. "Blocks" means we believe Google refuses the product; "at risk" means it serves but something is likely to go wrong or be dropped; "underperforming" means it serves but reaches fewer queries than it could. Where Google's own documentation states the consequence, we cite it; where it does not, we do not present a guess as a fact.

Why did the scan refuse to read part of my file?

Because the alternative is worse. If a file is ambiguous in a way that would move data between columns, the wrong delimiter putting prices into the availability field, for instance, every finding after that point would be fiction. In that case the scan stops and says so, rather than hand you a confident report about the wrong data. Where rows are merely dropped, the count states plainly how many, so a partial read is never presented as a complete one.

My export uses semicolons, or a non-UTF-8 encoding. Is that a problem?

No. A semicolon-separated file, which is what a European Excel produces, is read correctly, and the output is comma-clean. Encodings other than UTF-8 are read and transcoded, and a byte order mark is stripped before anything downstream sees it. These are mentioned in the report so the column mapping makes sense, not because you have to act on them.

Why is my valid-looking barcode rejected?

Usually a restricted prefix. A GTIN can have a correct check digit and still be refused because it begins with a range reserved for in-store or coupon use. Excel is the other frequent cause: it turns a long barcode into scientific notation and drops the leading zero. Both are covered on the identifiers page.

My pages rank in Google Search, so why are my Shopping ads refused?

Almost always robots.txt. The crawler that checks ad landing pages, AdsBot-Google, ignores the wildcard User-agent: * group that every other crawler obeys, so a rule that Search reads and works around is invisible to it. The whole mechanism, and the fix, is on the robots.txt page.

What date format does Google expect?

ISO 8601, for every date field, for example 2026-08-15T13:00+0200. The scan checks availability dates, expiration dates, and sale-price effective dates against it, and also catches an expiration date set more than 30 days ahead, which is the common thirty-versus-365 mistake.

Do you check EU requirements like unit pricing and energy labels?

Yes, and it is the part tools built for the US market tend to skip. Unit pricing where the price indication directive is likely to apply, the EPREL certification shape, and energy_efficiency_class restricted to the countries it is still valid for. These are flagged as prompts to confirm rather than firm verdicts, because a file cannot always tell whether a rule applies.

What about shipping and returns written into the feed?

Both are validated against the composite syntax Google actually documents, where the sub-attributes live in the column header and the values are positional. One point worth knowing: a single item-level shipping price can switch off your account-level shipping settings for that location, so the report raises a note whenever it sees one.

Can you tell me what Google actually decided about my products?

No. That needs access to your Merchant Center account, which we do not ask for. We can tell you what your data says; Google tells you what it did with it. The full list of what a file cannot establish is on what we don't check.

What happens to a product with no barcode at all?

Nothing, as long as you say so honestly. Some products genuinely have no GTIN, and that is declared with identifier_exists: no. What is caught is a contradiction: declaring that no identifier exists while also supplying a GTIN, an MPN, or a brand-and-MPN pair. The check does not invent a barcode for you, and it does not penalise a product for legitimately lacking one.

Two of my products share a barcode. Is that wrong?

It is worth knowing rather than automatically wrong. Each variant is meant to carry its own GTIN, so a repeated barcode across a group is flagged, but the message says why it is sometimes legitimate rather than the check trying to decide for you. This is the kind of finding we report and leave to your judgement rather than repair.

Will the scan reprice my catalogue if a value is ambiguous?

No, and this is deliberate. A value like 1.234 means one thousand two hundred and thirty-four in continental notation and just over one in English, and nothing in the file settles which. Rather than guess and risk repricing a product by a factor of a thousand, the scan reports it as ambiguous and leaves it for you to resolve.

Do you fix the file, or only report on it?

The scan reports; it never edits your file. If you buy a fix, the corrections are delivered as a separate supplemental file that Google reads alongside your feed, so your own export is never rewritten. What that means in practice is on how it works.

Do you check titles and descriptions?

Yes. Promotional text that gets a product disapproved, the word SALE, a percentage, a price in the title, is flagged, and a title over 150 characters is reported as a warning, because Google truncates it and serves the product with the title cut off rather than refusing it outright. Descriptions are checked for stray HTML, for the 5,000 character limit, and for being too thin for Google to understand what the product is.

Does it matter which country and currency I tell you?

Yes, and that is why you are asked at upload. A bare number in a price field cannot say whether it is euros, dollars or pounds, and guessing wrong would reprice a catalogue silently. The country you select also decides which regional rules the scan applies, since some requirements, unit pricing and energy labelling among them, depend on where you sell.

What is a supplemental data source, in plain terms?

A second file attached to your Merchant Center account that Google reads alongside your main feed. It cannot add or remove products; it only overrides named attributes on products your main feed already contains, matched on id. That is what lets a fix keep applying after your nightly export, and it is explained in full on how it works.

Your file and your data

Where does my file go, and for how long?

To our server, where it is scanned and then held for up to six hours, long enough for you to read the report and decide, because the repair needs the file again. After that it is deleted automatically. The corrections are not kept either: they go into a spreadsheet in your own account, and Merchant Center reads them from there. The full account is on privacy.

Can I delete my file sooner?

Yes. The report carries a control that deletes your upload from the server immediately, so you do not have to wait out the six hours or take our word for them.

Do you invent product information?

No. Every generated value is checked word by word against your own row, and anything introducing a claim your data does not contain is discarded and your original kept. Where a title is rewritten, that is declared to Google as generated content.

Who fetches my images?

Our server does, when it validates them, so it is our requests your image host sees rather than yours. Alongside the images, the scan reads your robots.txt over HTTPS; it does not crawl the rest of your site.

Is my data used to train an AI model?

The free scan sends your data to no model at all. Only the paid title rewrite involves one, and then only selected fields from the affected products, not the whole file, are sent to the model provider for that single step. What is sent, and what is not, is set out on privacy.