Google Images Scraper: image tiles and the page each one sits on, no key

Type a search and get one row per image tile: the Google thumbnail, the page hosting the image, that page's domain, the tile's title text and its rank in the results.

Read this before you buy, because it decides whether this is any use to you. Google stopped putting the original file address in the Images HTML, so there is no link to the full-size file here. Every row comes back with imageUrl: null and imageUrlAvailable: false rather than passing the thumbnail off as the original. What you get is thumbnailUrl and sourcePageUrl.

Input	One Google Images search
Output	One row per image tile
Ceiling	100 images per run, across up to 3 pages
Account needed	None, and no Google API key
Price	$0.40 per 1,000 images, flat on every plan

🖼️ What Google Images Scraper does

It runs your search against Google Images, reads the tile grid, and writes a row per tile.

Each row carries thumbnailUrl (the encrypted-tbn*.gstatic.com address Google serves in the grid), the thumbnail's width and height, sourcePageUrl and sourceDomain for the page the image sits on, the tile's title text, Google's own docId, and the rank and page it came from.

Set maxPages up to 3 to ask for more tiles. Duplicate tiles are removed across pages, so an image appearing twice arrives once.

📥 What you give it

{
  "query": "golden retriever puppy",
  "language": "en",
  "countryCode": "US",
  "maxItems": 20,
  "maxPages": 1
}

Field	Default	What it is
query	box starts at red running shoes	Your image search. Leave it empty and you get one uncharged sample row.
language	en	Google interface language, like en, fr or de.
countryCode	US	Two-letter country for localised results.
maxItems	20	Image rows to return, 1 to 100. It is also the most a run can bill you.
maxPages	1	Result pages to request, 1 to 3. More pages take longer.
safeSearch	active	Not applied. See below.
enableProxyFallback	on	Leave it on. Off means a refused first request has nowhere to go.
proxyConfiguration	leave it alone	Optional. Your own servers are used exactly as given, in the order listed.

safeSearch does nothing. It is still in the form, and every value behaves the same: Google returns a mostly unrelated tile set when the parameter is sent, so it is left off and you get Google's default filtering. Rows say so with safeSearchApplied: false. Do not build a content policy on that field.

📤 What you get back

A real row from a recent run:

{
  "ok": true,
  "rank": 1,
  "query": "golden retriever puppy",
  "page": 1,
  "title": "How to Train a Golden Retriever Puppy: Growth & Training Timeline",
  "imageUrl": null,
  "imageUrlAvailable": false,
  "thumbnailUrl": "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcRUNLBzUGVp8D2IE6U-7oqMxVJ0xJaE1GFBpikkO40bOA&s=10",
  "thumbnailWidth": 678,
  "thumbnailHeight": 452,
  "sourcePageUrl": "https://www.akc.org/expert-advice/dog-breeds/golden-retriever-puppy-training-timeline/",
  "sourceDomain": "www.akc.org",
  "width": null,
  "height": null,
  "docId": "jUS0xT_ulT-SUM",
  "safeSearchApplied": false
}

Field	What it is
thumbnailUrl	The only image address you get. It is Google's cached thumbnail, not the original file.
thumbnailWidth, thumbnailHeight	The served thumbnail's shape, read from the tile.
imageUrl, imageUrlAvailable	null and false on the current layout. The field exists so a row never pretends otherwise.
width, height	The original file's dimensions. Also null, for the same reason.
sourcePageUrl, sourceDomain	The page carrying the image. This is what most people actually want.
docId	Google's own tile identifier, handy as a dedupe key.
rank, page	Position within its page, and which page it came from.

🧾 Reading the output

Row	How to spot it	Charged
An image tile	ok: true and no _sample	yes
The sample row	_sample: true	no
A diagnostic	ok: false and an errorCode	no

The sample row is not representative of a real one. It shows imageUrl, width and height filled in, and real rows on the current layout never carry them. Run a one-image search to see the truth.

Code	What it means
NO_RESULTS	The page came back with no tiles on it. A very narrow search lands here.
BLOCKED	Google would not serve that request. Re-run it or reword the search.
RATE_LIMITED	Google answered with a rate limit. Wait and try again.
JS_REQUIRED	Google served a page that builds itself in the browser rather than a tile grid.
NETWORK	Google was unreachable or the response never finished.

The Overview table shows imageUrl, width and height, which are always empty, and hides thumbnailUrl and sourceDomain, which are not. Read the dataset as JSON or export it, rather than judging a run by that table.

▶️ How to run it

1. Open Google Images Scraper and click Try for free.
2. Put your search into Image search query, replacing the example.
3. Set Maximum images. Twenty is a sensible first run.
4. Change Language and Country if you want another market's results, then click Start.
5. Download the dataset as JSON, CSV or Excel, or read it from the Apify API.

💰 How much does it cost?

$0.40 per 1,000 images, which is $0.0004 a row. The same rate on every Apify plan, no volume tiers.

maxItems caps the rows, so it also caps what a run can bill: set it to 20 and twenty rows is the most you can be charged for.

Not charged as images: the sample row, every diagnostic row, duplicate tiles removed across pages, and a search that comes back with nothing.

💡 What people use it for

• Finding which sites are using a product photo, by reading sourceDomain across the whole result set.
• Building a shortlist of pages to visit for licensing, where the source page matters more than the file.
• Comparing what one search returns in US against DE or JP, tile for tile.
• Checking how your own pages rank in image results for a keyword, using rank and sourceDomain.

🚧 What it does not do

• No original file addresses. imageUrl, width and height are null on every row from the current layout. If your project needs the full-size file, open sourcePageUrl and find it there.
• safeSearch is not applied, whatever you set it to. safeSearchApplied: false on every row.
• 100 images and 3 pages per run. There is no way to ask for the thousandth result.
• A failure on page two or three throws away the earlier rows. With maxPages above 1, a run that trips on a later page writes a diagnostic and no image rows. Keep it at 1 when one page is better than nothing.
• One search per run. Use a schedule or several runs for a keyword list.
• It does not download images. You get addresses and metadata, and what you do with them is yours to get right: licences, copyright and each site's terms are your responsibility.
• Rows are a snapshot. Image results reshuffle constantly.

🧭 Which Google scraper do you need?

If you want	Use
Image tiles and their source pages	This one
The ordinary web results page for a keyword	Google Search Results Scraper
News articles with the publisher's real link	Google News Scraper
Job postings from the jobs box	Google Jobs Scraper
Businesses, places and their reviews	Google Maps Scraper

❓ Questions people ask

Can I get the full-size image file? Not from Google. The address is not in the HTML any more. Open sourcePageUrl and take it from the page itself.

Why is width empty when thumbnailWidth is filled? thumbnailWidth is the tile Google served. width would be the original file, and Google no longer tells anyone what that is.

Does safeSearch filter anything? No. Every row carries safeSearchApplied: false.

How many images can one search return? Up to 100, across at most three pages. Google often runs out before that on a narrow search.

Can I schedule it? Yes. Runs are independent, and docId makes deduping across runs easy.

Is scraping Google Images legal? The rows are public search metadata and addresses. The images themselves belong to whoever made them, so check licences and each site's terms before reusing anything. Apify's write-up on scraping and the law is a good starting point, and we are not lawyers.

🆘 If something breaks

Open the Issues tab on the actor page. Send the search and the run ID. The errorCode on the diagnostic row usually names the problem on its own.