Google Ads Transparency Scraper with Ad Text

Every ad an advertiser, a domain or a keyword runs on Google, straight from the official Google Ads Transparency Center, with the words of the ad: headline, description, display URL, and for local businesses the name, street address, city, rating and phone the ad shows.

Most Transparency Center scrapers stop at a link to the ad preview. This one reads the preview for you, and when Google only keeps a text ad as a picture (it does for most big advertisers), it reads the picture with OCR. You get a spreadsheet of ad copy, not a list of links.

See it run

Who uses it

• PPC and marketing agencies: find local businesses that are already paying for Google Ads in a niche and a country, with what they say in their ads. A business that runs ads has a budget.
• Competitor research: every headline and description a competitor has run, with first and last shown dates and days on air. Copy that runs for months is copy that converts.
• Brand protection and compliance: who is bidding on your brand or sending ads to your domain.
• Lead generation for local services: roofers, dentists, plumbers, insurance agents advertising in the US right now, with their website and, on Maps ads, their address and rating.

What you get, measured

Measured on 8 October 2026 in two runs on the Apify platform (keywords "roofing" and "dentist", domains geico.com and angi.com, United States):

	Result
Text ads with a headline	93 of 95 (98 %)
Text ads with a description	85 of 95 (89 %)
Text ads with the display URL	73 of 95 (77 %); Google Maps ads show a business name instead of a URL
Text ads read from the HTML preview	17
Text ads read with OCR (Google keeps them only as an image)	78, average OCR confidence 85-96 %
Speed and platform cost	108 ads in 94 seconds, about $0.0001 per ad of platform usage

Said plainly:

• Image (banner) ads come with the image URL, not with text: reading a graphic design with OCR gives loose words, not a headline, and the Actor does not pretend otherwise.
• Video ads come with the YouTube link.
• OCR is OCR: about 4 % of text ads read from an image carry a stray character where a photo touches the text. The field textSource says ocr or html_preview, and ocrConfidence gives the score, so you can filter.
• Fields Google does not show in a given ad are absent from that row, never invented.

Input

• Keywords: like the Transparency Center search box ("roofing", "dentist miami"). Each keyword brings the matching advertisers; the run takes those from your country first, then the ones with most ads (maxAdvertisersPerQuery, 5 by default).
• Domains: every ad that sends traffic to a website, whoever paid for it.
• Advertiser IDs: the AR... ID from a Transparency Center URL.
• Country: two-letter code (US, GB, CA, AU, DE, FR, ES, MX, BR, IN and 25 more) or anywhere.
• Formats: text, image, video, or all.
• Only ads shown in the last N days: 7 or 30 gives what is running now.
• Max ads per advertiser or domain and Max ads in total.
• Read the ad text (on by default): turn off for links only.
• Only ads new since the last run: remembers the ads already delivered in a named key-value store in your account. Schedule it weekly to watch competitors and pay only for new ads.

Output

One row per ad:

{
  "advertiserId": "AR14331040638106075137",
  "advertiserName": "Roofing PD",
  "creativeId": "CR06225560223742951425",
  "format": "text",
  "headline": "Michigan's Roofing Experts",
  "description": "Roofing, Siding, Gutters, & More From A Certified, Licensed Roofing Company! Call Today",
  "displayUrl": "roofingpd.com",
  "domain": "roofingpd.com",
  "businessName": "Roofing PD",
  "addressStreet": "5073 Canterbury Dr",
  "addressCity": "Brighton",
  "addressRegion": "MI",
  "addressPostalCode": "48114",
  "addressCountry": "US",
  "firstShown": "2025-01-17T19:27:41.000Z",
  "lastShown": "2026-10-07T12:01:52.000Z",
  "daysShown": 385,
  "lastShownDaysAgo": 0,
  "textSource": "html_preview",
  "hasText": true,
  "adUrl": "https://adstransparency.google.com/advertiser/AR14331040638106075137/creative/CR06225560223742951425?region=US",
  "regionFilter": "US",
  "query": "roofing",
  "scrapedAt": "2026-10-08T09:12:45.000Z"
}

Other fields when the ad has them: adLayout (search or maps), rating, reviewCount, category, services, phone (E.164), extraTexts (sitelinks and other lines), adTexts (every text of the ad in order), ocrConfidence, imageUrl, previewUrl, youtubeVideoId, youtubeUrl.

RUN_SUMMARY in the key-value store lists, per advertiser or domain, the ads read and delivered and how many came with text, plus the totals read from previews and by OCR.

Pricing

Pay per ad delivered (event ad), plus a small Actor start fee. Ads filtered out, and ads already delivered when "only new ads" is on, are not charged. The text is included in the price of the ad.

Reliability and fair use

• The Actor asks the Transparency Center the same questions its own website asks, one request at a time with a pause between them, and backs off when Google says to slow down.
• It does not try to get past Google's "unusual traffic" check. If Google answers with its CAPTCHA page, the run stops, delivers what it already has and says so in the log and in RUN_SUMMARY (blockedByGoogle). Run it again later.
• An advertiser or domain that fails does not stop the others.
• Data: public ad transparency records published by Google about advertisers (companies). No personal data beyond what an advertiser puts in its own ad.

Release test

npm test runs two checks: test/parse.js reads real ads saved on 8 October 2026 (a local search ad, a search ad with a URL path, a video ad and a text ad Google keeps as an image), and test/smoke.js asks the live Transparency Center (keyword search, two pages of one advertiser, the country filter, a preview and an OCR read). From some home IP addresses Google answers the live test with its CAPTCHA; the test then says so instead of trying to pass it, and has to be run from Apify.