How to Compare Public Google Ad Creatives Across Regions

Setup and public build details

Google Ads Transparency public build 0.2.7 accepts targets containing names, domains, advertiser IDs, or public advertiser/creative URLs; region, platform, adFormat, optional dateFrom/dateTo, includeDeliveryDetails, and maxItems. A saved Task preset uses targets such as `["nike.com", "adidas.com"]`, region `CA`, platform `ALL`, adFormat `ALL`, includeDeliveryDetails `false`, and maxItems `10`. Replace the sample targets with one intended checked advertiser ID for your first check: targets `[oneID]`, region `CA`, platform `ALL`, adFormat `ALL`, includeDeliveryDetails `false`, and maxItems `2`. Use the exact native adFormat, not an invented input. Optional dateFrom/dateTo settings must agree across compared samples; do not invent pagination.

Required native fields include input, inputType, requestedRegion, requestedPlatform, requestedFormat, advertiserId, advertiserName, advertiserLegalName, advertiserDomain, advertiserCountryCode, advertiserVerified, creativeId, creativeUrl, observedAt, status, and statusReason. Optional payerName, servedRegions, and servedPlatforms can be null. Keep requestedRegion separate from advertiserCountryCode and optional delivery geography; do not infer targeting, audience nationality, or private performance.

OUTPUT represents run evidence containing targetsReceived, targetsResolved, targetOutcomes, duplicatesSkipped, unusableCreativesSkipped, creativesSaved, statusCounts, and stopReason. Actual current bounded acceptance differs from fiction: an exact-ID Canada pilot returned one useful creative under cap 2, with requestedRegion ca, advertiserCountryCode SG, advertiserDomain null, servedRegions null, and servedPlatforms null with delivery details off; null is unavailable, not no delivery. OUTPUT reported one target received and resolved, one saved creative, status ok, and sources_finished. A cap of 2 is a ceiling, not a promised number, and sources_finished does not certify complete country coverage. Only Canada was tested; a real second-country comparison requires an accepted matching second execution before comparing. Do not present US fiction as product acceptance.

Steps for regional creative comparison

1. Confirm the exact advertiser ID first. A domain name can match multiple accounts, so check account inclusion before you start.
2. Choose two countries. Keep the platform, ad format, date settings, and item caps identical for both runs. Keep the advertiser country separate from your requested ad region.
3. Run a small test first. For Canada, use targets with your known advertiser ID, region CA, platform ALL, ad format ALL, includeDeliveryDetails false, and maxItems 2. A cap is a limit, not a promised count.
4. Save each native creative row. Note the advertiser ID, creative ID, creative URL, region, and status. Do not guess audience targeting or performance from these dates.
5. Check each source run before you compare sets. A failed query or hit cap is a source limit, not an empty population. If inputs, identity, or collection dates differ materially, label the samples unmatched and do not attribute differences to the region filter.
6. Build sets using the combined advertiser ID and creative ID key for each region sample. Report intersection and each sample-only set using that exact key. Unknown IDs remain unresolved. Multiple variations in one creative are not multiple creative IDs.
7. Describe results clearly. Use labels like observed in both, region A only, region B only, or not seen in this sample. Keep the latter distinct from never delivered. Result differences may reflect source selection, caps, time changes, or missing details rather than proven advertiser strategy.
8. Deliver your membership table and receipt. If a query fails or an advertiser mismatch remains, stop the comparison and keep your single-region evidence. Do not combine an anywhere sample with a country sample.

Original-fiction regional sample ledger and receipt

The following original-fiction sample receipt and membership table illustrate a reader-made receipt, not native fields or an actual US product run. Suppose advertiser A has a matched-input receipt collected under identical platform ALL, ad format ALL, cap 5, includeDeliveryDetails false, explicit dateFrom 2026-01-01 and dateTo 2026-03-31, with concrete observation timestamps on 2026-02-15 for both Canada and the United States. Materially different advertiser settings or collection boundaries create unmatched samples, not a region effect.

Below is the original-fiction five-key compound advertiserId plus creativeId regional membership table. Here, AC1, AC2, AC3, AC4, and AC5 represent fictional creative keys for advertiser A.

In this original-fiction set, the Canada sample size is 3, the United States sample size is 4, the intersection size is 2 (AC2, AC3), Canada-only size is 1 (AC1), United States-only size is 2 (AC4, AC5), and the union size is 5. "Not seen" means absent only within that saved checked sample, not never delivered. Duplicate observations of creative C3 contribute only one known key, and unknown keys stay unresolved outside trusted sets. If the United States execution instead fails, that set is unknown rather than empty, and exclusive membership arithmetic cannot be computed from it.