Google Images Scraper

Scrape Google Images into structured rows. Every image comes back with its full-resolution URL, its exact dimensions, the page it appears on and that page's title, the domain, the file size Google measured, and the image's dominant colour. Where Google recognises the image as a product, the row also carries the price, the list price, the star rating, the review count and the brand.

✅ What you get

• The source page, not just the image. Page URL, page title, site name and domain on every row.
• File size and dominant colour — fileSize as Google reports it (14KB, 1.2MB), dominantColor as an rgb() triple.
• Product data on shopping images. Price, list price, currency, star rating, review count, brand, description and the Google Shopping link.
• Exact dimensions for both the full image and the thumbnail.
• A provenance link (aboutThisImageUrl) to Google's "About this image" page.
• Google's own filters: size, colour, type, file type, usage rights, recency, SafeSearch, single-domain, and 18 languages × 20 countries — each one checked against a live baseline, not assumed.
• A working aspect-ratio filter, applied here rather than by Google (see the limits below).
• Partial-failure rows, so one blocked search does not destroy a bulk run.

❌ What this isn't

This is not a reverse image search, an image downloader, or Google Lens. It does not fetch the image files themselves — it returns the URLs and the metadata around them.

🔎 Example tasks

• Product images with prices and ratings
• Creative Commons images, large only
• Images from one domain

👥 Who it is for

SEO and content teams auditing which pages rank in image search; e-commerce teams watching how their products appear next to competitors'; designers and researchers building filtered image sets; anyone assembling a training or reference set who needs attribution, not just pixels.

⚙️ Input

Give it queries (one search term per line), or paste searchUrls straight from Google Images — a pasted URL keeps whatever filters you already set there. Everything else is optional: size, colour, type, aspect ratio, file type, usage rights, published-within, a single domain, SafeSearch, language and country. The prefilled input is a complete one-search run.

📤 Output

One row per image:

Field	Example
imageUrl, imageWidth, imageHeight	https://…/shoe.jpg, 1872, 2340
thumbnailUrl, thumbnailWidth, thumbnailHeight	https://encrypted-tbn0.gstatic.com/…, 225, 225
fileSize	14KB
dominantColor	rgb(42,48,48)
sourcePageUrl, sourcePageTitle	the page the image is published on, and its <title>
sourceDomain, siteName	www.adidas.com, Adidas
aboutThisImageUrl	Google's provenance page for the image
isProduct	true on shopping results
productTitle, productBrand, productPrice, productListPrice, productCurrency	199.99, was 300, USD
productRating, productReviewCount, productDescription	4.59, 81
googleShoppingUrl	the product's Google Shopping page
query, searchUrl, position, scrapedAt	provenance for the row itself

💵 Pricing

Pay per image returned. No start fee, no charge for a search that comes back empty.

⚠️ Honest limits

• 100 images per search is Google's ceiling, not ours. There is no page two on Google Images: &start=100 answers 200 with an empty result set. To go wider on one topic, run the same term under different filters (size, colour, recency) — each combination returns a different 100.
• Product data covers roughly one row in eight. That is the share Google itself tags as shopping results; the rest are ordinary web images and their commerce columns are left empty rather than guessed.
• Google ignores its own aspect-ratio filter on image search, so this Actor applies it. In a controlled run, asking Google for tall images returned 98 of the same 100 images as the unfiltered search and only 22 of them were actually portrait. Every row carries exact dimensions, so the filter is applied to the results instead — which means a narrow shape returns fewer than 100 rows, and you are only charged for the rows you get. Every other filter was confirmed to bite against the same baseline.
• "Exactly WxH" is roughly four rows in five. Google fills the remainder with near matches; the dimensions on each row tell you which is which.
• usageRights reports what Google reports. It reflects what the source site claims, it is not legal advice, and you should check the licence on the source page before reusing an image.
• Results are personalised by locale. The same query under a different country or language returns a different set, and product prices follow the country you choose.
• A blocked search fails loudly. A run where no search returned any image ends as FAILED rather than shipping an empty dataset as a success.