Google Images Scraper
Pricing
from $1.00 / 1,000 image scrapeds
Google Images Scraper
Google Images search results as JSON: original image URL, thumbnail, exact pixel dimensions, title, source page and domain for every image. No browser, no API key. From $1.50 per 1,000 images.
Pricing
from $1.00 / 1,000 image scrapeds
Rating
0.0
(0)
Developer
subimpact
Maintained by CommunityActor stats
0
Bookmarked
1
Total users
1
Monthly active users
14 hours ago
Last modified
Categories
Features
Share
Turn Google Images into clean JSON. Give it one or more queries and get one row per image: the direct link to the original full-size image, Google's thumbnail, the exact pixel dimensions, the title, the page the image sits on, its domain and the site name — no browser, no API key, no scraping code to maintain.
- $1.50 per 1,000 images, no start fee. You pay only for image rows that are delivered. Empty queries and failed queries cost nothing.
- Pure HTTP against Google's Images view, parsed from the page itself, so it is fast (about 5-8 s per page of ~100 images) and cheap to run.
- Country (
gl), interface language (hl), SafeSearch and Google's own size / type / colour / time filters are supported. - Every row carries a
status, so a run that hit a block or an empty query explains itself instead of silently returning nothing.
What you get
One dataset item per image:
| Field | Meaning |
|---|---|
query | The query this image was found for. |
position | 1-based rank within the query, in Google's order, across pages and after duplicates are dropped. |
title | The title Google shows (usually the page title). |
imageUrl | Original image file URL on the source site — not a Google thumbnail. |
thumbnailUrl | Google's own thumbnail on encrypted-tbn0.gstatic.com (small, fast, always reachable). |
width, height | Pixel size of the original image, as Google reports it. |
sourcePageUrl | The page the image appears on (where Google's card links). |
sourceDomain | That page's domain, e.g. en.wikipedia.org. |
sourceName | The site name Google displays, e.g. Wikipedia, Instagram. |
scrapedAt | ISO 8601 timestamp of the fetch. |
status | ok for an image row, empty when Google had no images for the query, error when the query failed. |
error | Reason for an empty/error row, otherwise null. |
Input
{"queries": ["red panda", "mid century modern sofa"],"gl": "us","hl": "en","maxImagesPerQuery": 50,"maxItems": 200,"safeSearch": true,"imageSize": "large","imageType": "photo","colour": "","timeRange": ""}
Only queries is required. A default run (one query, 50 images) costs about $0.08.
| Input | Default | Notes |
|---|---|---|
queries | — | Up to 50 queries. Duplicates and blank lines are dropped. |
gl | us | Google country. Changes the result set and the regional mix of sites. |
hl | en | Google interface language, e.g. en, de, pt-BR. |
maxImagesPerQuery | 50 | 1-500. Google serves ~100 images per page, so 50 is one page. |
maxItems | 200 | Run-wide cost ceiling across all queries. |
safeSearch | true | Sends safe=active. |
imageSize | any | large, medium, icon — Google's own size filter. |
imageType | any | photo, clipart, lineart, animated. |
colour | any | red, orange, yellow, green, teal, blue, purple, pink, white, gray, black, brown, blackandwhite, transparent. |
timeRange | any | day, week, month, year. |
concurrency | 3 | Queries fetched in parallel (1-8). Pages of one query are always sequential. |
Filters
All four filters are Google's own: they are sent as Google's tbs parameter and applied by Google's
index, so the result set is exactly what a human sees after clicking the filter. For example
imageSize: "large" roughly halves the number of results and raises the smallest short side from 320 px
to 674 px, imageType: "animated" returns mostly .gif originals, and colour: "red" vs "blue"
share almost no results. Leave a filter empty for "any".
Output example
A real row from a live run of this actor (queries: ["red panda"]):
{"query": "red panda","position": 1,"title": "Red Panda - SAFE Worldwide","imageUrl": "https://safeworldwide.org/wp-content/uploads/2022/05/red-panda-2022-03-05-09-19-19-utc-scaled-e1653444292143-768x740.jpg","thumbnailUrl": "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcT2_Z64mpT-r7D5_ueGW8FVDsUJ79nkKg-VzC4byVeW_g&s=10","width": 768,"height": 740,"sourcePageUrl": "https://safeworldwide.org/red-panda/","sourceDomain": "safeworldwide.org","sourceName": "SAFE Worldwide","scrapedAt": "2026-11-14T19:36:50.377Z","status": "ok","error": null}
And a query with no images (delivered, not charged):
{"query": "a query google has no images for","position": null,"title": null,"imageUrl": null,"thumbnailUrl": null,"width": null,"height": null,"sourcePageUrl": null,"sourceDomain": null,"sourceName": null,"scrapedAt": "2026-11-14T19:36:50.377Z","status": "empty","error": "Google returned no images for this query"}
How many images can one query return?
Google's Images view serves about 100 images per page. This Actor walks Google's own result offset
(start=0, 20, 40, ...) and keeps going until it reaches maxImagesPerQuery or Google runs out.
Measured on 2026-11-15: page 1 and page 2 of red panda returned 100 images each with no overlap, and a
deeper offset returned Google's "there aren't any Images matches on this topic" page. For cat, pages 1-4
returned 100, 100, 46 and 49 images before Google stopped. So expect up to a few hundred images per
query — maxImagesPerQuery above that is harmless, the run just ends when Google has nothing more.
If you need more coverage, add more queries: that is how Google's Images index divides up.
Cost
$1.50 per 1,000 images — about $0.0015 per image — with no monthly or start fee, roughly half of what
the most popular paid alternative charges for the same data. You are charged one image-scraped event
per delivered image row. Empty and error rows are free, a failed push is free, and the run stops by
itself at maxItems so a query you typed wrong cannot run up a bill.
Notes, limits and honest caveats
- Original image links can rot or block hotlinking.
imageUrlpoints at the source site's own file (that is what "original" means). Some hosts refuse requests without a referer, restrict hotlinking, or delete the file later.thumbnailUrlis Google's cached copy and is the more stable of the two; both are returned so you can fall back. - Dimensions are Google's numbers. They come from Google's own metadata for the indexed image, which matched the real files in every check we made — but if the source site replaces the file, the numbers can drift from what you download.
- Titles are Google's titles. For social posts Google often shows the post text instead of a page title; the row is faithful to Google, not normalised.
- Google's depth is finite. There is no page 3 for every query (see above), and a few very broad queries return fewer results than their page count suggests.
- Blocking is possible and handled, not eliminated. The actor fetches through Apify's Google SERP
proxy (included in the price; you do not need your own proxies), retries a blocked page with a fresh
proxy session, and reports a query that never got through as an
errorrow that is not charged. In normal testing we saw no CAPTCHA pages on this view, but Google can serve them to a tired proxy IP. - SafeSearch is on by default. Turn it off only deliberately.
- Results are Google's ranking for the
gl/hlyou asked for, so the same query ingl=usandgl=myreturns different sites.
Run monitoring and privacy
If a run fails or times out, this Actor tells its developer the run ID and status, so we can fix problems fast. It never sends your input, your results or your account data.