Google Ads Transparency Scraper: Ad Library API

Every ad a company runs on Google Search, YouTube, Shopping, Maps and Play, read from Google's public Ads Transparency Center: headline, copy, call to action, images or YouTube video, landing page, first and last shown dates, and for ads shown in the EU the impressions per country and audience targeting. Search by advertiser name, website domain, advertiser ID or URL. No login, no browser, no API key.

Try it: Click Start with the default input (nike.com, 20 most recently shown ads): 20 ads in 7 seconds for about $0.01 (run on October 21, 2026). Each ad comes with its advertiser, format, dates, days shown and a link to the ad on the Transparency Center; text ads that Google stores as screenshots come with the screenshot URL.

Run on Apify on October 14, 2026 (19:51 UTC). Input: grammarly.com, region Germany, platform YouTube, format video, 8 ads, per-country details on. It finished in 18 seconds.

#	Headline	CTA	YouTube video	First shown	Days shown	Countries	EU impressions
1	Ideas to essays	Download	JQX7SdUJGeo	2026-03-25	205	22	125K–150K
2	Write with purpose, work with pride	Download	G6wkI_fhsCY	2026-03-25	205	18	20K–25K
3	Edit. Get set. Send.	Download	G6wkI_fhsCY	2026-05-28	139	12	under 1K
4	Work leveled up	Download	G6wkI_fhsCY	2026-03-25	200	22	300K–350K
5	Think. Write. Fix. Done.	Download	QoI7QEvAljo	2026-03-25	205	22	1M–1.25M
6	Check originality	Download	VBAQr43v5hc	2026-03-25	205	22	225K–250K
7	Instant citations	Download	daGUD3CB7UU	2026-03-25	152	9	under 1K
8	Edit. Get set. Send.	Download	uN1N6VCtQvY	2026-05-27	140	12	under 1K

Pricing: $0.50 per 1,000 ads, plus $0.50 per 1,000 for the optional per-country details add-on and $0.50 per 1,000 advertiser rows; $0.00005 per run start. Platform usage is included. Failed runs, error rows and ads skipped by the watcher are not charged.

Real example in full

Row 5 of the table above (per-country list cut to 3 of 22, variations to 2 of 5):

{
  "type": "ad",
  "advertiserId": "AR05288677867637440513",
  "advertiserName": "Superhuman Platform Inc.",
  "creativeId": "CR17624851031638671361",
  "format": "video",
  "targetDomain": "grammarly.com",
  "firstShown": "2026-03-25",
  "lastShown": "2026-10-14",
  "daysShown": 205,
  "adUrl": "https://adstransparency.google.com/advertiser/AR05288677867637440513/creative/CR17624851031638671361?region=DE",
  "adType": "display_template",
  "headline": "Think. Write. Fix. Done.",
  "description": "Watch how Humanizer makes AI text sound authentic and human.",
  "headlines": ["Think. Write. Fix. Done.", "Humanizer transforms AI text into writing that feels natural and authentic"],
  "displayUrl": "grammarly.com",
  "landingUrl": "https://www.grammarly.com/a",
  "callToAction": "Download",
  "imageUrl": "https://tpc.googlesyndication.com/simgad/7959286080260022750",
  "videoId": "QoI7QEvAljo",
  "videoUrl": "https://www.youtube.com/watch?v=QoI7QEvAljo",
  "videoDuration": "1:05",
  "channelName": "Grammarly",
  "paidBy": "Superhuman Platform Inc.",
  "regionCount": 22,
  "impressionsMin": 1000000,
  "impressionsMax": 1250000,
  "platformImpressions": [{"platform": "YouTube", "impressionsMin": 1000000, "impressionsMax": 1250000}],
  "regions": [
    {"region": "DE", "regionName": "Germany", "firstShown": "2026-03-25", "lastShown": "2026-10-14", "impressionsMin": 250000, "impressionsMax": 300000},
    {"region": "SE", "regionName": "Sweden", "firstShown": "2026-03-25", "lastShown": "2026-10-14", "impressionsMin": 225000, "impressionsMax": 250000},
    {"region": "IE", "regionName": "Ireland", "firstShown": "2026-03-25", "lastShown": "2026-10-14", "impressionsMin": 200000, "impressionsMax": 225000}
  ],
  "audienceSelection": {"demographicInfo": "included", "geographicLocations": "included and excluded", "contextualSignals": "included and excluded", "topicsOfInterest": "included and excluded"},
  "variationCount": 5,
  "variations": [
    {"headline": "Think. Write. Fix. Done.", "callToAction": "Download", "videoId": "QoI7QEvAljo"},
    {"headline": "Sound like you", "callToAction": "Download", "videoId": "QoI7QEvAljo"}
  ],
  "region": "DE",
  "platform": "YouTube",
  "input": "grammarly.com",
  "scrapedAt": "2026-10-14T19:51:07Z"
}

Other runs the same day:

• Load test: temu.com in the US, 500 ads with ad content, in 2 minutes 24 seconds (312 requests, 2 retries). The 500 ads were 182 video, 162 image and 156 text ads.
• Who advertises a domain (advertisers mode): the 300 most recent ads pointing to temu.com came from two advertisers, ELEMENTARY INNOVATION PTE. LTD. (280 ads: 169 text, 68 video, 43 image) and Whaleco Inc. (20 ads).
• Watcher: Booking.com in the UK, shown in the last 2 days, 40 ads. The first run returned 40 ads; the second run returned none, skipped the 40 it already knew and charged $0.

What you can search

Put one item per line in searches. You can mix kinds in one run.

You enter	Example	You get
Website domain	nike.com	Every ad that points to the domain, from any advertiser (brands, agencies, resellers)
Advertiser name	Nike, Inc. or Nike	Ads of the advertiser with that name and the most ads (or all same-name advertisers, see Use every advertiser with the same name)
Advertiser ID	AR16735076323512287233	All ads of that advertiser
Transparency Center URL	https://adstransparency.google.com/advertiser/AR…/creative/CR…	That ad (or that advertiser, or that domain search)

Filters: region (any country, or anywhere), platform (Search, YouTube, Shopping, Maps, Play), ad format (text, image, video), and a date window (shown in the last N days, or between two dates). Results come most recently shown first. Max ads per input caps how many ads are read per line.

Modes

• Ads (default): one row per ad, as in the example.
• Advertisers: find advertiser IDs. For a name you get every advertiser Google suggests, with registered country, ad count range and whether the name matches exactly. For a domain you get every advertiser running ads that point to it, with how many of the scanned ads are theirs and their formats.

Watching for new ads

Turn on Only ads not seen before, give the task a watcher name, set Shown in the last N days to 2 or 3 and schedule the actor daily. Every run returns only ads that earlier runs under the same watcher name did not return, and skipped ads cost nothing. The memory is kept in the named key-value store google-ads-transparency-state. Use firstShown to tell an ad that really just launched from one that was running before the watcher first saw it.

{"searches": ["gymshark.com"], "region": "GB", "lastDays": 3, "maxAdsPerSearch": 100,
 "onlyNew": true, "watcherName": "gymshark-uk-new-ads"}

Set Max ads per input above the number of ads the advertiser shows in that window; the run log prints it, as in gymshark.com: about 64-64 ads match. The watcher reads only the newest ads up to that cap, so for a very large advertiser it returns ads it simply had not read yet. On 2026-10-16, booking.com showed 100,000 to 200,000 ads in the UK over two days. A second run capped at 100 ads returned 100 ads the first run had not read, and the newest of them was first shown in August 2026. Gymshark showed 64 ads over the same two days, so a 200-ad cap covers all of them and the watcher returns only new ones.

A real catch: a daily watcher on gymshark.com in the UK (lastDays: 2, maxAdsPerSearch: 200, watcher name readme-gymshark) started on 2026-10-16 with 64 ads. Its run at 07:21 UTC on 2026-10-19 returned 2 ads, skipped the 69 it already knew, and charged for the 2. One was a YouTube ad launched the day before:

{
  "type": "ad",
  "advertiserName": "Gymshark Ltd",
  "creativeId": "CR01419556965582897153",
  "format": "video",
  "targetDomain": "gymshark.com",
  "firstShown": "2026-10-18",
  "lastShown": "2026-10-18",
  "daysShown": 1,
  "videoUrl": "https://www.youtube.com/watch?v=2O4ncz2TQh0",
  "region": "GB",
  "adUrl": "https://adstransparency.google.com/advertiser/AR05501100765344694273/creative/CR01419556965582897153?region=GB"
}

The other was an image ad first shown on 2026-05-12 that had just come back into the 2-day window. That is why firstShown matters: filter on it to keep only launches.

Pricing

Pay per event. Platform usage (compute, proxy, storage) is included.

Event	Price	When
Ad	$0.50 per 1,000	One ad row, including headline, copy, images and video where Google has them
Ad details add-on	$0.50 per 1,000	Only with Per-country stats on: countries, EU impressions, targeting, variations
Advertiser	$0.50 per 1,000	One advertiser row in advertisers mode
Actor start	$0.00005	Once per run

Examples: all 500 recent temu.com ads cost $0.25. 1,000 ads with per-country stats cost $1.00. A daily watcher on 10 competitors that finds 30 new ads a day costs about $0.45 a month. Error rows, "no results" rows and ads skipped by the watcher are never charged. The six most used Google Ads Transparency actors charge $1.00 to $2.00 per 1,000 ads on the free plan, and one charges $3.00 per 1,000 for details on top.

Accuracy and limits

• Most text ads are stored as pictures. For most Search text ads Google only keeps a screenshot of the ad: in our tests about two thirds of temu.com's text ads and all of booking.com's and nike.com's (US) text ads; in the default nike.com run on October 21, 2026, 13 of 20 ads. Those rows have adType: "archived_image" and an imageUrl of the screenshot, but no headline text. Turning on Per-country stats sometimes recovers the text from the ad's other variations.
• Impressions and targeting are EU only. Google publishes impressions ranges and audience selection under the EU Digital Services Act, so they exist only for ads shown in EU countries. Elsewhere impressionsMin/impressionsMax are null. Impressions are ranges (e.g. 1M–1.25M), as on the website, and Google updates them with a delay.
• Some previews do not render. When Google cannot render an ad's preview either, the row says adType: "preview_unavailable" and keeps the dates, IDs and links. Video ads sometimes come with only a YouTube thumbnail in the list (video_thumbnail); turn on Per-country stats to read the full ad from its variations.
• No spend data. Google publishes spend only for political ads, which this actor does not cover.
• Names are matched loosely by Google. The advertiser-name search ignores legal suffixes (Inc, Ltd, BV...). If the wrong advertiser comes back, run advertisers mode first and use the ID.
• Landing URLs come from the ad preview. Search text ads show the display URL but not the landing URL.
• Date window means the ad was shown at least once in the window; it can have started earlier.
• Big advertisers have a lot of ads (Google lists 9,000–10,000 for Nike, Inc. and 200,000–300,000 ads point to temu.com in the US), so set Max ads per input.

FAQ

Is this data public? Yes. Google publishes it in the Ads Transparency Center for anyone to view, without signing in. The actor reads the same data the website shows.

Can I use it from my code or an AI agent? Yes. Call it through the Apify API or the Apify MCP server. A run for one domain and 20 ads takes about 7 seconds.

How do I find an advertiser's ID? Run advertisers mode with the brand name or domain, or copy it from the advertiser's page URL on adstransparency.google.com (it starts with AR).

Why does a domain search show agencies? Agencies and resellers that run ads pointing to a brand's site appear under their own advertiser names, which is often exactly what competitor research needs.

Does it need a proxy? No. Requests go directly from the Apify server. If Google rate limits them, the actor retries and moves to the Apify datacenter proxy automatically.

Watcher scheduling: use a separate watcherName for each independent workflow and keep its filters/location consistent. Runs using the same name wait up to 50 seconds for the previous run; if it is still busy, retry after it finishes. Local and cloud baselines are separate. A failed or interrupted delivery can appear again on retry, so deduplicate downstream by creative ID and observation time. Existing names keep their baselines, but a baseline previously shared by colliding names cannot be disentangled automatically.

Changelog

• 0.1.11 (2026-10-21): README restructured: default-input timing, pricing line and a ready-to-schedule watcher input at the top. No code changes.
• 0.1.10 (2026-10-19): Watcher runs with the same name now wait for each other before loading and saving their baseline. Losing the lock stops the run without saving stale state. Cancelled runs keep only confirmed output in watcher state.
• 0.1.9 (2026-10-19): Spending-limit fix: a row is charged and remembered only when it is delivered; add-ons are charged only with their row; watcher names with special characters no longer collide.
• 0.1.8 (2026-10-19): README shows a real watcher run that caught a newly launched ad. No code changes.
• 0.1 (2026-10-14): First release. Ads and advertisers modes, region, platform, format and date filters, ad content (headline, copy, images, video, app info, sitelinks), per-country details with EU impressions and audience selection, new-ads watcher.