Google Ads Transparency Scraper: ads, text and dates by advertiser

Collect ads from Google's Ads Transparency Center by advertiser name, website or advertiser ID. Get a screenshot or preview link, first and last shown dates, days shown, available ad text, variations and country details. No Google account or browser is needed.

One price per ad covers available text, details, compute and proxies (see Pricing). Turn on Return only ads not collected before to avoid collecting and paying for the same ads again.

How often you get ad text: Google stores many ads only as a screenshot or video, so no scraper can read words from those; you still get a screenshot, image or preview link for every ad. Where Google keeps the text, this scraper reads it. On the same 12 Nike ads in the US it returned text for 10, as many as the most-used alternative scraper, while another popular one returns no ad text at all. On 30 Shopping product ads from three US retailers it returned the product title for 22, the most-used alternative scraper for 20 on the same search. Across our test runs, 40% of all ads and 85% of US ads came with readable text (see Limits for other countries).

What it returns

• Advertiser name and ID, creative ID and a link to the ad on Google.
• Ad format: text, image or video.
• First and last shown dates, plus Google's number of days shown.
• A screenshot/image URL and/or a Google preview URL. Video records also get a YouTube watch link when Google exposes an unambiguous video ID.
• Headline, description and displayed website where Google provides readable text, and the product title and store of Shopping product ads. Text availability varies greatly; see Limits.
• Available rendering variations, country-level first/last shown dates and impression ranges. Details are on by default; turn them off for speed.
• Or, in advertiser summary mode, one row per advertiser: who advertises for a website or name, their country, verification, legal name, approximate ad count and latest ad.

Input

Field	What it does	Default
queries	Advertiser names, websites such as nike.com, or IDs starting with AR	example: nike.com
resultType	ads (one row per ad) or advertisers (one row per advertiser, see Advertiser summary mode)	ads
maxAdvertisersPerQuery	Advertiser mode only: advertisers per entry, 1-50	5
region	Country where ads were shown, such as US, GB, DE or IN; empty means anywhere	anywhere
platform	SEARCH, YOUTUBE, SHOPPING, MAPS or PLAY; empty includes all	all
format	TEXT, IMAGE or VIDEO; empty includes all	all
dateFrom, dateTo	Optional shown-between dates, YYYY-MM-DD; the end date is inclusive	no date filter
maxAdsPerQuery	Maximum ads per entry, most recently shown first; in monitoring mode only unseen ads count	40
includeAdText	Read available headlines, descriptions, displayed websites and Shopping product titles	on
includeDetails	Collect variations and available country dates/impression ranges	on
onlyNewAds	Remember collected creative IDs and skip them in later runs	off
proxyConfiguration	Residential proxy for advertiser searches, ads and details; no country is forced. Text previews normally load directly, with one proxy fallback if blocked	residential

Websites give the most reliable results. For a name, the actor picks one advertiser from Google's own name suggestions. It prefers names that start with your words (Flipkart finds Flipkart Internet Private Limited), then advertisers registered in your selected country, then the advertiser with the most ads. Google's suggestions do not always contain the company you mean: Canva can offer only Canvasso and similar names. Use canva.com or the advertiser ID (AR..., shown in the ad's Transparency Center link) instead. The run's RUN_SUMMARY record lists the chosen advertiser and the top five candidates for each name. A website can return ads from several advertisers or agencies.

Example: collect up to 100 Nike ads shown in the US:

{
  "queries": ["nike.com"],
  "region": "US",
  "maxAdsPerQuery": 100
}

Example: monitor a competitor's new Google Search ads in the US (schedule it daily; each run returns only ads you have not collected before):

{
  "queries": ["shopify.com"],
  "region": "US",
  "platform": "SEARCH",
  "maxAdsPerQuery": 30,
  "onlyNewAds": true
}

A run collects roughly 2,500-3,500 ads per hour with text and details on (50 ads take about a minute). The default run timeout is 3 hours; for more, raise Timeout in the run options or split the entries over several runs.

Use it from code

Replace <YOUR_API_TOKEN> with your token from Apify Console > Settings > API & Integrations.

Python (pip install apify-client):

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("cozybit/google-ads-transparency-scraper").call(
    run_input={"queries": ["nike.com"], "region": "US", "maxAdsPerQuery": 20}
)
for ad in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(ad["advertiser_name"], ad["format"], ad["last_shown"], ad["image_url"])

JavaScript (npm install apify-client):

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<YOUR_API_TOKEN>' });
const run = await client.actor('cozybit/google-ads-transparency-scraper')
    .call({ queries: ['nike.com'], region: 'US', maxAdsPerQuery: 20 });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.length, items[0]);

HTTP API, one call that waits for the run and returns the ads (for runs under 5 minutes):

curl -X POST "https://api.apify.com/v2/acts/cozybit~google-ads-transparency-scraper/run-sync-get-dataset-items" \
  -H "Authorization: Bearer <YOUR_API_TOKEN>" -H "Content-Type: application/json" \
  -d '{"queries": ["nike.com"], "region": "US", "maxAdsPerQuery": 20}'

Apify can also run this on a schedule and send results to webhooks, Google Sheets, Zapier or Make through its integrations.

Output (record format v1)

One record per collected ad. You can download the dataset as JSON, CSV or Excel. Example with includeDetails: false and includeAdText: false:

{
  "query": "nike.com",
  "input_type": "domain",
  "advertiser_id": "AR16735076323512287233",
  "advertiser_name": "Nike, Inc.",
  "creative_id": "CR14091266735822864385",
  "format": "TEXT",
  "first_shown": "2023-04-25T20:05:00.205Z",
  "last_shown": "2026-10-30T16:24:49.587Z",
  "days_shown": 1108,
  "image_url": "https://tpc.googlesyndication.com/archive/simgad/1095568144160320364",
  "preview_url": null,
  "video_url": null,
  "ad_url": "https://adstransparency.google.com/advertiser/AR16735076323512287233/creative/CR14091266735822864385?region=US",
  "headline": null,
  "description": null,
  "display_url": null,
  "ad_text_status": "unavailable",
  "region": "US",
  "status": "ok",
  "error": null,
  "scraped_at": "2026-10-30T17:00:00.000Z"
}

With details enabled (the default), records also contain variations: [{image_url, preview_url}] and regions: [{geo_target_id, first_shown, last_shown, impressions_min, impressions_max}]. Unavailable values are null; impression bounds are strings to preserve precision. Country dates use YYYY-MM-DD; overall first/last shown timestamps are UTC. days_shown is Google's served-day count, not a date subtraction or an impression count.

ad_text_status is ok when text was extracted, not_text for image ads without readable text, or unavailable when readable copy was not exposed, extraction was disabled or optional enrichment failed. A valid ad remains status: "ok" and is charged even when text/details are unavailable; error explains an enrichment failure if there was one. A query that fails before finding an ad produces no fake ad row. Query failures, empty searches, selected advertisers, requests, skipped IDs and missing visuals are recorded in the run's RUN_SUMMARY key-value-store record.

Advertiser summary mode

Set resultType to advertisers to answer "who advertises for this website, and how much?" without collecting every ad. Each entry returns up to maxAdvertisersPerQuery rows:

• Website: the distinct advertisers behind the 200 most recently shown ads for that website (with your filters), in order of appearance. A website can be advertised by the brand, its regional companies, agencies and resellers.
• Name: the best matches from Google's name suggestions, ranked as described in Input.
• Advertiser ID: that advertiser.

Field	Meaning
advertiser_name, advertiser_id, advertiser_url	Advertiser and its page on Google
advertiser_country, legal_name, verified	Country and legal name from Google's advertiser record; verified is Google's identity verification (null if unknown)
ad_count_min, ad_count_max	Google's approximate number of this advertiser's ads matching your filters; for a website entry, only ads for that website
query_ad_count_min, query_ad_count_max	Website entries: approximate ads for the website from all advertisers
latest_ad_last_shown, latest_ad_format, latest_ad_url	The advertiser's most recently shown matching ad

Example: who advertises for three websites in the US:

{
  "queries": ["nike.com", "booking.com", "shopify.com"],
  "region": "US",
  "resultType": "advertisers"
}

This returned 11 advertisers in 60 seconds, the same 11 as the most-used alternative scraper's advertiser mode, including Nike's Dutch and Korean companies and a media agency. An entry without ads returns one row with status: "no_data". A known advertiser without matching ads keeps its row with counts of 0. If Google's advertiser record or count fails, the row stays with those fields empty and an error. onlyNewAds, includeAdText, includeDetails and maxAdsPerQuery do not apply in this mode. Ad counts are Google's ranges, not exact numbers.

Pricing

Trial price until 30 November 2026: 0.01 USD per 1,000 ads (the lowest price Apify allows). From 1 December 2026 this scraper costs 1.00 USD per 1,000 ads. Apify notifies all users before a price change.

Event	Price
One ad record, including available text, details, dates and links	1.00 USD per 1,000 (0.001 USD each)

• 100 collected ads cost 0.10 USD.
• 1,000 collected ads cost 1.00 USD.
• Previously seen ads skipped in monitoring mode cost nothing. Empty queries also have no ad charge.
• Advertiser summary rows currently have no charge.
• Text or details being unavailable does not make the ad free: the delivered ad, dates and visual links are the charged record.

There is no separate text/details add-on or additional proxy/compute charge. Set a maximum cost per run in the run options to cap spending; no new work starts once it is reached, and ads not delivered are neither charged nor marked collected.

Free Apify plan

Free-plan runs return at most 50 ads (or 50 advertiser rows) in total, shared across all entries, even if you request more. At the limit the run finishes successfully with: "Free Apify plan limit reached: up to 50 ads per run. Upgrade to any paid Apify plan to remove it." Everything already collected remains in the dataset. Any paid Apify plan removes this actor-imposed cap.

Monitoring mode

Turn on onlyNewAds and reuse the same entry, country, platform and format on later runs. The first run collects normally. Later runs skip previously collected creative IDs before looking up details/text or charging. The maximum per entry counts new ads, not skipped ones.

History is stored in the named key-value store google-ads-transparency-state, separately for each normalized entry/country/platform/format combination. It remembers the latest 50,000 collected IDs per combination and drops the oldest beyond that limit. Only successfully emitted records enter history; unpushed records do not. Changing a date filter or the text/details options does not reset history. Use one run at a time for the same combination to avoid concurrent updates overwriting history. Deleting a history record resets that combination, so later collections can include and charge those ads again.

Later searches look at ads shown since three days before the previous run started, unless your dateFrom is later. The three days cover Google reporting serving dates late; an ad that Google reports more than three days late can be missed. The first run still obeys your requested maximum.

Reliability

In 16 test runs on the Apify platform with default settings (7 countries and "anywhere"; names, websites and advertiser IDs; mixed platform and format filters), every run finished without a failed entry and all 264 ads came back as valid records, each with a screenshot, image or preview link. Google blocked 4 requests along the way; the actor recovered every time on a new IP. The median run took 30 seconds.

How it stays reliable: Google blocks shared datacenter IPs, so searches and ad details go through residential IPs, each making one slow, paced request at a time. A blocked IP or a failed proxy connection is replaced, up to four times per entry.

Limits

• Readable ad text is not guaranteed. Share of ads with text in our test runs: US 85%, Brazil 63%, Germany 25%, India 16%, UK none in our sample. Image ads usually keep their words inside the picture; Shopping product ads come with their product title.
• Screenshots are returned as image_url. No OCR or video transcription is used. Video ads get their companion text when Google includes it, plus a YouTube link.
• Keep the default residential proxy. Datacenter proxies are not recommended because Google blocks shared datacenter IPs.
• Dates, variants and impression ranges are returned only where Google exposes them. region filters where an ad was shown, not necessarily where its advertiser is registered.
• A name finds a best match from Google's suggestions, not a guaranteed identity (see Input). Prefer websites or advertiser IDs.
• A maximum number of ads is not a guarantee that that many exist for your filters. Google can omit, delay or remove ads and can change its undocumented endpoints.
• Preview URLs and media URLs can expire. Save the data or download permitted media promptly if you need long-term access.
• Google's public transparency data is the source; this actor is not affiliated with Google.

Support

Use the actor's Issues tab for problems and feature requests. Include the run link, the entry and filters you used, and the creative ID for a missing field. Do not include proxy passwords or API tokens.