Google Ads Transparency Center Scraper

Export public advertisers and advertising creatives from Google Ads Transparency Center using anonymous HTTP requests. This actor uses Python 3.12, the Apify SDK, httpx, and Beautiful Soup. It does not launch a browser or require a Google account. Its internal RPC adapter was checked against real responses on September 26, 2026. These endpoints are undocumented and can change without notice. Review VALIDATION.md before deployment, especially the unsupported Display platform filter.

Quick start

Create a Python 3.12 virtual environment inside this directory and install requirements.txt. Run python -m unittest discover -s tests -v for offline checks. Run apify validate-schema .actor/input_schema.json to validate the Console input form. To exercise the actual SDK entry point with isolated local storage, run powershell -File validation/run_live.ps1 -Python .venv/Scripts/python.exe. The harness stops a hung subprocess after 115 seconds and saves results under validation. It excludes inherited Apify credentials from the child environment.

The supplied INPUT.json searches nike.com and Shopify, restricts serving evidence to the United States, and caps output at 40 creatives per input. It explicitly disables Apify Proxy so the daily smoke test needs no keys. The measured default run took approximately 23 seconds and returned 80 creatives. Normal hosted inputs default to Apify Proxy enabled; configure proxy access in the hosting account before using that default.

Inputs and matching

advertisers accepts AR identifiers, advertiser names, and domains. AR identifiers receive an advertiser lookup. Names use SearchSuggestions and select the matching suggestion with the largest estimated creative count. This is a documented heuristic, not proof of company identity: inspect the returned advertiser name and ID, then use an explicit ID for repeatable monitoring. Domain searches can match several regional legal entities. They return a separate advertiser row for every resolved ID encountered among matching creatives.

region accepts a supported ISO alpha-2 country code or anywhere, the default. The checked-in country mapping comes from Google's public frontend configuration. formats is a multiselect of text, image, and video; all three are enabled by default. platforms optionally selects YouTube, Search, Maps, or Shopping. The current RPC enum has no Display value. Selecting Display produces an uncharged warning and skips that selection; selecting only Display performs no searches. An unfiltered search can still return image creatives.

dateFrom and dateTo accept inclusive YYYY-MM-DD boundaries on the creative's last-shown date in UTC. Filtering happens locally against returned timestamps; records without usable dates are excluded when a date filter is present. maxCreativesPerAdvertiser defaults to 200 and is applied as a combined cap per submitted input, including domain searches spanning multiple advertiser IDs. Duplicate advertiser and creative records are suppressed across the entire run. Sparse filtered searches also have a finite page budget and report truncation with a summary row.

includeCreativeDetails, default false, fetches each selected creative's detail endpoint and extracts rendering variations and available country/date information. timeoutSecs bounds each HTTP operation. proxyConfiguration accepts the standard Apify proxy object. No credentials belong in this repository or input examples.

Dataset contents

Advertiser rows contain rowType, original input, advertiserId, advertiserName, verified, region, domains, creativeCountEstimate, url, and source. Verification inverts Google's unverified flag. Domain lists contain observed input domains, not an exhaustive advertiser portfolio. Estimates are source ranges or null; they are not exact inventory counts.

Creative rows contain advertiserId, advertiserName, creativeId, format, platforms, regions, firstShown, lastShown, previewUrl, landingDomain, text, youtubeVideoId, url, and source. Optional details add variations. Missing information stays null or empty. Platforms and regions represent known evidence, not exhaustive coverage: without a platform filter the platform list is empty, and overlapping filtered discoveries retain the first observation. Preview URLs are rendered image URLs or identified YouTube thumbnails, never renderer JavaScript links. Some formats supply neither. Text extraction uses visible renderer markup; image OCR is outside this actor's scope. Preview links may expire.

Reliability and pricing

Requests retry twice with one- and two-second backoff for HTTP 429, server errors, or transport failures. Proxy sessions rotate every eight requests. A redirect to /sorry retries through up to three new proxy sessions, then stops with an uncharged warning. Without a proxy, a block stops immediately. No CAPTCHA solving or signed-in session access is attempted.

Each unique successful advertiser row requests one advertiser-lookup event at $0.01. Each creative row requests one creative-returned event at $0.002. Summary and error rows are free. Local non-PPE runs do not bill. Configure both events from .actor/pay_per_event.json and disable synthetic events before publication. Charging precedes persistence to respect charge limits; a storage failure after charging cannot be rolled back atomically. This project has not been pushed or published.

Platform validation update

The earlier local-only and no-push statements above describe the original build. Claude's 2026-09-26 platform handoff in the project MESSAGES.md records an immediate /sorry block without proxy: one summary, zero creatives. Apify datacenter proxy returned four advertisers and 80 creatives in 30 seconds. This actor needs Apify datacenter proxy on the platform; it is now on by default, including INPUT.json. The earlier statement that supplied INPUT disables proxy is historical. This observed success does not guarantee future access.

Example output

One recorded dataset row, trimmed by omitting fields only. Values are the saved snapshot, not current measurements. Source: validation/results-default.json, first row in the rows array.

{
  "rowType": "advertiser",
  "advertiserId": "AR18378488041124659201",
  "advertiserName": "Nike Retail BV",
  "verified": true,
  "region": "NL",
  "domains": [
    "nike.com"
  ],
  "creativeCountEstimate": null
}

Use cases

• A paid media agency can collect competitor creatives by explicit advertiser ID and assemble a dated review board of available previews.
• A brand team can resolve a domain to regional advertiser entities and manually confirm identities before repeating creative monitoring.
• A creative strategist can filter video or image records and inspect available text and links when preparing a campaign research workshop.
• A market analyst can compare creatives returned under different region filters while retaining each run as a separate evidence snapshot.

Pricing example

Hypothetical batch, calculated from .actor/pay_per_event.json:

Event	Count	USD per event	Subtotal
creative-returned	1,000	$0.002	$2.0000
advertiser-lookup	10	$0.01	$0.1000

Total declared event charges: $2.10. These counts are a budgeting example, not a promised yield or an actual bill. Any applicable platform or proxy costs are outside this calculation.

Limitations

Display-specific filtering is unavailable. Returned creatives do not establish spend, conversions, or effectiveness. Null previews and incomplete platform evidence do not prove that advertising is absent.