Google Images Scraper: image URLs, thumbnails, titles and source pages

Search Google Images by keyword, country and language, and get one row per image: the image URL, the thumbnail, the title, the page it appears on, the domain and the position in Google's results. Run as many searches as you like in one go, apply Google's own filters (size, type, color, format, license, date), and set a hard limit on images per search and in total.

No browser to maintain and nothing to configure: you give search terms, you get images.

What you get

For every image:

Field	What it is
query, queryIndex	The search term and its position in your list (0-based)
position	Rank of the image in Google's results for this search (starts at 1)
imageUrl, directImage	Address of the full image on its original site. directImage is false for Instagram and Facebook posts, where Google gives an address that opens the post page rather than the file
thumbnailUrl	Google's small cached copy
title, alt	Title Google shows; alternative text when it differs
sourceName	Name of the site that hosts the page (the domain instead, when the site name is a person's name)
pageUrl, domain	Page where the image appears, and its domain (no www.)
imageDomain, format	Domain serving the file; jpg, png, gif, webp, svg... read from the address (null when it does not say)
country, language, scrapedAt	What was asked, and when

Each run also writes an OUTPUT record with, for every search (the first 200), the number of images delivered, Google's related searches and spelling suggestion, and why the run stopped.

Example row

{
  "type": "image",
  "status": "ok",
  "query": "mountain lake",
  "queryIndex": 0,
  "position": 2,
  "imageUrl": "https://www.nps.gov/im/nccn/images/NOCA_MOLA_LowerEast.jpg",
  "thumbnailUrl": "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcSy0XJ24ERC3ss3x6OOSMkahXIysmwpQnmZyPBmImkw5w&s=10",
  "title": "Mountain Lakes Monitoring (U.S. National Park Service)",
  "alt": null,
  "sourceName": "National Park Service",
  "pageUrl": "https://www.nps.gov/im/nccn/mountain-lakes.htm",
  "domain": "nps.gov",
  "imageDomain": "nps.gov",
  "format": "jpg",
  "directImage": true,
  "country": "US",
  "language": "en",
  "scrapedAt": "2026-10-22T05:21:14.000Z"
}

Input

Field	Meaning
queries	Search terms, one per line, up to 5,000. A Google Images link is read as its search term. Operators such as site: and filetype: work (see Pricing).
maxResultsPerQuery	Images per search, 1 to 200 (default 100). Google serves about 100 per search page and at most 200 per term.
maxItems	Stop after this many images in total (optional).
countryCode, languageCode	Where and in which language the search is made (US/en by default). Codes or names.
imageSize, minResolution	Large, medium, icon; or "at least 400 x 300 ... 70 MP"
imageType	Photo, clip art, line drawing, animated (GIF), face
color	Full color, black and white, transparent, or one color
fileType	jpg, png, gif, webp, svg, bmp, avif, ico, raw
aspectRatio	Tall, square, wide, panoramic
usageRights	Creative Commons licenses, or commercial and other licenses
uploadedWithin	Past 24 hours, week, month, year
safeSearch	Turn Google SafeSearch on

{
  "queries": ["mountain lake", "alpine cabin"],
  "maxResultsPerQuery": 100,
  "countryCode": "US",
  "languageCode": "en",
  "imageType": "photo",
  "minResolution": "4mp"
}

Google returns different images for each filter combination, so repeating a search with another filter is the way to collect more than 200 images for one subject.

Pricing

Pay per event: a small fee per run, a fee per search Google answers, and a price per image delivered. Nothing else is charged: no fee for duplicates, none for error rows.

Event	Free and Bronze	Silver	Gold and above
Run start	$0.005	$0.005	$0.005
Search	$0.003	$0.003	$0.003
Image delivered	$0.0030	$0.0029	$0.0028

• Search: charged once per search term that Google answers, even when it has no image. A search with an operator such as site:, filetype:, intitle:, inurl: or related: is charged as 5 searches, which is what it costs to run; quotes, -word, before: and imagesize: are charged as one. Each search is charged at most once per run.
• Image: $3.00 per 1,000 images on the base plan, $2.90 on Silver and $2.80 on Gold.
• The run start is charged once per run; the platform may scale it with memory (this Actor runs on 256 MB).

Example: 50 images from one search cost $0.158 on the base plan. 5,000 images (50 searches of 100) cost $15.155.

Your maximum charge per run is never exceeded: a search starts only when it and at least one image fit in what is left, and the run stops when the next image would exceed it, and says so.

Limits, said plainly

• Up to 200 images per search term. That is the most Google lists for one search. Use filters or related terms for more.
• No pixel size per image. Google's result list does not carry width and height, so the fields are not there rather than guessed. Use minResolution to guarantee a minimum.
• Not the full web. Positions are what Google shows to a visitor from the chosen country at the time of the run; they change over time and between runs.
• Image rights belong to their owners. The Actor collects links and metadata; check each image's license before reusing it.
• Duplicates are removed per search (same image address). A search that returns fewer images than asked delivers and charges only those images, plus the search.
• The first results of a new search term can take a few seconds; a run of 50 searches takes about a minute.

FAQ

Is anything different on Apify's free plan? Available on paid Apify plans only, except for the prefilled example (the search "mountain lake", 20 images, US, English, no filter), which returns stored sample rows from the day the Actor was built, free of charge, on any plan. Any other input on a free plan returns a notice and charges nothing but the run start. Reason: Apify does not pay developers for runs on free plans, and each search has a real cost.

What happens if a search finds nothing? You get a free error row with the reason and the other searches carry on; the run does not fail because one search is empty. no_results (Google answered with no image) and search_failed (Google could not run it, for example site: with a site it does not know) are charged as a search, since Google was asked; no image is charged. A search term refused before searching (invalid_query) or a search the service could not answer (no_answer, rejected_query) is not charged at all, unless Google answered an earlier try of that search: then it is charged once as a search, never for images.

What if I stop or restart the run? Images already delivered are kept and are not charged again, and a search already made is not made or charged again. A run moved to another server resumes where it stopped. A run that had finished does nothing more if you resurrect it.

Can I search in another country or language? Yes: countryCode and languageCode accept codes (FR, fr, pt-BR) or names. Results are those a visitor from that country sees.

Does the Actor send anything to Quarra? Only when a run fails, runs out of time, or stops early because the image service is unavailable: the Actor name and build, the type of error, the line of our code where it happened, and the memory and time limit of the run. Never your input, your results, the error text or your account.

Is it legal to collect this data? The Actor reads the public result list of Google Images and does not log in. Check the terms of the sites you reuse images from, and the license of every image you publish.

Support

Open an issue in the Issues tab: we reply within 48 hours and tell you what was fixed and in which version. — Sam, Quarra