Google Images Scraper - Image URLs, Sizes & Source Pages

Type a search phrase and get the Google Images results as a table: direct image URL, width and height, thumbnail, title, source page and domain, in Google's order. No login and no API key. A run with the prefilled query red bicycle and 20 images takes about a minute and costs about $0.06 on the Free plan.

Use it to collect image links for ML datasets, to check which sites rank in Google Images for a keyword, or to find visual references. The Actor saves links only: it does not download the images.

Table of contents

• What it does
• Quick start
• Input
• Output
• Pricing and limits
• Use cases
• API, export and integrations
• Troubleshooting
• FAQ
• Support

What it does

1. Searches Google Images for each phrase in queries, in the country and language you choose.
2. Applies the Google filters you set: SafeSearch, size, type, color, time period and usage rights.
3. Reads up to 3 Google result pages per phrase. One page has about 100 images, so Google gives about 300 images per phrase at most.
4. Removes repeated images (the same image URL) inside one phrase and numbers the rest 1, 2, 3… in Google's order.
5. Saves one Dataset row per image and stops a phrase when it has maxImages images or when Google has no more.

What data you get

For every image: the search phrase, position, page title, full-size image URL, width and height in pixels, Google thumbnail URL and size, source page URL, domain and site name, file size, main color, the country and language used, and the time Google's page was loaded.

Good fit

• Build an image dataset for machine learning: many phrases, up to about 300 image links each, with sizes to filter small images.
• SEO: see which pages and domains show up in Google Images for your keywords, and at which position.
• Design and e-commerce: collect visual references, or check where your product photos appear.

Not a fit / limits

• It does not download or store images. You get links; the image files stay on the source sites and can disappear.
• It does not give image licences. The usageRights filter is Google's label, not a licence. Check the rights on the source page before you use an image.
• About 300 images per phrase at most, because Google shows no more. For more images, use more phrases (for example red bicycle vintage, red bicycle kids).
• No reverse image search (search by picture) and no Google Lens.
• Google results change over time and between runs. Two runs of the same phrase can differ by about 10% of images.

Quick start

1. Open the Input tab.
2. Keep the prefilled phrase red bicycle in Search queries (queries), or type your own, one per line.
3. Keep Images per query (maxImages) at 20, or choose 1–300.
4. Click Start.
5. Open the Output tab and look at the Images view: a preview, the title, the image URL, the size and the source page for each image.

Expected result: 20 rows for red bicycle in about a minute, positions 1 to 20.

Input

Input fields

Only queries is required. Every other field has a default.

Field	Plain name	Type	Required	Default / prefill	Allowed values	Effect	Recommendation
queries	Search queries	array of strings	yes	prefill ["red bicycle"]	1–100 phrases, up to 500 characters each	One Google Images search per phrase	Empty lines are ignored; repeated phrases (case-insensitive) run once. Google operators such as site: and quotes work.
maxImages	Images per query	integer	no	default 20, prefill 20	1–300	How many images to save for each phrase	Up to 100 needs one Google page per phrase; 101–300 needs up to 3.
countryCode	Country	string	no	us	two-letter code from the list, e.g. us, gb, de, fr, jp	Google ranks results for searchers in this country	Use your target market.
languageCode	Language	string	no	en	code from the list, e.g. en, de, fr, ja, pt-BR	Google interface language; changes ranking and the language of titles	Match the language of your phrases.
safeSearch	SafeSearch	boolean	no	true	true, false	true hides explicit images	Keep true unless you need unfiltered results.
size	Image size	string	no	any	any, large, medium, icon	Google's size filter	large for datasets that need big images.
type	Image type	string	no	any	any, photo, clipart, lineart, animated, face	Google's type filter	photo removes most drawings and logos.
color	Color	string	no	any	any, color, gray, transparent, red, orange, yellow, green, teal, blue, purple, pink, white, black, brown	Google's color filter	transparent for PNG cut-outs.
timePeriod	Time period	string	no	any	any, day, week, month, year	Only images Google indexed in this period	Use for fresh images on a news topic.
usageRights	Usage rights	string	no	any	any, creativeCommons, commercial	Google's usage-rights filter	It is a Google label, not a licence.

If queries is missing or has only empty lines, the run fails at once with Invalid input; only the run start ($0.005) is charged.

Interacting settings

When	Input A	Input B	Effective behavior	Cost/result effect
Many phrases	queries	maxImages	maxImages is per phrase, not per run	10 phrases × 20 images = up to 200 rows
Strict filters	size, type, color, timePeriod, usageRights	maxImages	All filters are combined; Google can have fewer images than maxImages	The phrase ends as exhausted or not_found; you pay for pages read
Phrase in another language	queries	countryCode, languageCode	Google uses all three	Titles come in the language Google picks for that page

Input examples

Prefill: one phrase, 20 images

{
  "queries": ["red bicycle"],
  "maxImages": 20
}

Expected behavior: one Google page is read, 20 images are saved. Expected output: 20 rows with position 1–20, imageUrl, width, height, sourceDomain. Cost note: 1 start + 1 page + 20 images.

Dataset: several phrases, large photos, Germany

{
  "queries": ["Fahrrad rot", "Fahrrad Kinder", "Rennrad"],
  "maxImages": 300,
  "countryCode": "de",
  "languageCode": "de",
  "size": "large",
  "type": "photo"
}

Expected behavior: up to 3 Google pages per phrase, about 280–300 images per phrase. Expected output: up to 900 rows; each phrase has its own positions from 1. Cost note: 1 start + up to 9 pages + one charge per saved image.

Output

One Dataset row = one image. The Images view shows a short table with a preview; the All fields view shows every field. A run summary is saved as RUN_SUMMARY in the key-value store.

Output fields

Field	Meaning	Type	Can be empty?	Example
query	The search phrase	string	no	red bicycle
position	Position in Google Images for this phrase, from 1, after repeated images are removed	integer	no	1
title	Title of the page the image comes from	string	yes	24 in. Red 21-Speed … Mountain Bike … - The Home Depot
imageUrl	Direct URL of the full-size image on the source site	string	no	https://images.thdstatic.com/…_600.jpg
width	Full-size image width, px	integer	yes	600
height	Full-size image height, px	integer	yes	600
thumbnailUrl	Google thumbnail URL (encrypted-tbn*.gstatic.com)	string	yes	https://encrypted-tbn0.gstatic.com/images?q=tbn:…
thumbnailWidth	Thumbnail width, px	integer	yes	554
thumbnailHeight	Thumbnail height, px	integer	yes	554
sourcePageUrl	Page that shows the image	string	yes	https://www.homedepot.com/p/…/325010151
sourceDomain	Domain of the source page	string	yes	www.homedepot.com
sourceName	Site name as Google shows it	string	yes	The Home Depot
fileSize	File size as Google shows it (text, not a number)	string	yes	39KB
dominantColor	Main color of the image	string	yes	rgb(216,66,69)
countryCode	Google country used	string	no	us
languageCode	Google language used	string	no	en
fetchedAt	When the Google page was loaded (ISO 8601, UTC)	string	no	2026-10-15T04:51:03.022Z

null means Google did not show this value for the image. It is never replaced with 0 or an empty string.

Output example

One real row from the prefill run (red bicycle, 20 images), with long URLs shortened:

{
  "query": "red bicycle",
  "position": 1,
  "title": "24 in. Red 21-Speed Mechanical Disc Brake Mountain Bike for Teenagers outwyadironch43 - The Home Depot",
  "imageUrl": "https://images.thdstatic.com/productImages/ee2aa086-bc34-4653-8fda-ab7879d28920/svn/reds-pinks-bikes-outwyadironch43-64_600.jpg",
  "width": 600,
  "height": 600,
  "thumbnailUrl": "https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcT5OkkXYwisrBprY_ohTFjwk4QbVwfWv62tA9h93MKQGg&s=10",
  "thumbnailWidth": 554,
  "thumbnailHeight": 554,
  "sourcePageUrl": "https://www.homedepot.com/p/24-in-Red-21-Speed-Mechanical-Disc-Brake-Mountain-Bike-for-Teenagers-outwyadironch43/325010151",
  "sourceDomain": "www.homedepot.com",
  "sourceName": "The Home Depot",
  "fileSize": "39KB",
  "dominantColor": "rgb(216,66,69)",
  "countryCode": "us",
  "languageCode": "en",
  "fetchedAt": "2026-10-15T04:51:03.022Z"
}

Run summary and statuses

RUN_SUMMARY has the run status, a message, and one entry per phrase with its own status, the number of images and pages.

Phrase status	Meaning	Rows in Dataset
complete	Got maxImages images	maxImages
exhausted	Google had fewer images than maxImages	all Google had
not_found	Google has no images for this phrase	none
error	Google could not be read after 3 tries	rows saved before the error
stopped	The run reached your spending limit or timeout while this phrase was running	rows saved before the stop
skipped	The run stopped before this phrase started	none

Run status	When	Run result
complete	Every phrase is complete, exhausted or not_found	Succeeded
partial	Spending limit, timeout or abort stopped the run; or some phrases had error	Succeeded at a limit; Failed when a phrase had error (saved rows stay)
error	Google could not be read for any phrase	Failed
invalid_input	Input is wrong	Failed, only the run start is charged

Pricing and limits

Pricing unit

The Actor uses pay-per-event pricing. You pay only for these events; platform usage is included.

Event	When it is charged	Free plan	Bronze	Silver	Gold+
apify-actor-start	once per run	$0.005	$0.005	$0.005	$0.005
search-page	per Google result page read (about 100 images), also when Google shows no images	$0.01	$0.01	$0.01	$0.01
image	per saved image row	$0.002	$0.0018	$0.0016	$0.0014

A phrase that Google has no images for costs one search-page and saves no rows. A Google page that could not be read after 3 tries is not charged. Invalid input costs only the run start ($0.005). If your spending limit is reached, the run stops as partial and keeps only the rows you paid for.

Cost examples

Free-plan prices:

Scenario	Input scope	Events	Charge	Stop/cap behavior
Prefill	1 phrase × 20	1 start + 1 page + 20 images	$0.055	stops at 20
10 images	1 phrase × 10	1 start + 1 page + 10 images	$0.035	stops at 10
100 images	1 phrase × 100	1 start + 1–2 pages + 100 images	$0.215–0.225	stops at 100
Full phrase	1 phrase × 300	1 start + 3 pages + about 290 images	about $0.62	Google gives no more
1,000 images	10 phrases × 100	1 start + 10–20 pages + 1,000 images	$2.11–2.21	stops at 100 per phrase
No images	1 phrase that Google does not know	1 start + 1 page	$0.015	phrase not_found

Limits

• 1–100 search phrases per run, 1–300 images per phrase. Google shows about 280–295 unique images per phrase, so a phrase can end with fewer than you asked for (status exhausted).
• Measured on the cloud (256 MB): 20 images ≈ 45–70 s; 10 phrases × 100 images ≈ 2–3 min; 30 phrases × 300 ≈ 11 min; 100 phrases × 300 (28,568 images) ≈ 34 min.
• The default timeout is 1 hour, which fits the largest input. Memory is fixed at 256 MB; peak use was about 80 MB.
• Each Google page gets up to 3 attempts and up to 120 seconds. A phrase whose pages keep failing gets status error; you are not charged for failed pages.
• If Google cannot be read on the first phrase, the run stops there and does not start the other phrases (status skipped). Later in a run, 3 failed pages in a row also stop new phrases.
• The Actor saves image links, not image files.

Use cases

Buyer job	Input path	Output entity	Key fields	Next integration
ML image dataset	many queries, maxImages 300, size large	image row	imageUrl, width, height, query	download the URLs with your own tool, export CSV/JSON
SEO in Google Images	your keywords, countryCode	image row	position, sourceDomain, sourcePageUrl	schedule weekly, compare positions in a sheet
Visual references	a few phrases, type, color	image row	thumbnailUrl, imageUrl, sourcePageUrl	open the Images view, export to Google Sheets

Scheduling and monitoring

Create a schedule in Apify Console (for example weekly) with the same input. Each run is a full new snapshot; the Actor does not compare runs. To see changes, compare position and sourceDomain between two Datasets.

API, export and integrations

• API: start the run with POST https://api.apify.com/v2/acts/cliqtomedia~google-images-scraper/runs and the input JSON; read rows from the run's default Dataset. See the Apify API reference.
• Export: JSON, CSV, Excel, XML or HTML from the Output tab or the Dataset API.
• Scheduling: Apify schedules with the same input.
• Webhooks / no-code: Apify integrations, Make, Zapier or n8n run the Actor and read the Dataset rows.

Use with AI agents (MCP)

• MCP server URL with only this Actor: https://mcp.apify.com?tools=cliqtomedia/google-images-scraper (Apify MCP server docs).
• Smallest useful input: {"queries": ["red bicycle"], "maxImages": 20}.
• Key output fields: imageUrl (full-size image link), width and height (pixels), sourcePageUrl (page with the image), position (rank in Google Images), title (page title).
• A run with 1 phrase and 20 images takes about a minute and costs $0.055 on the Free plan. Up to 100 phrases and 300 images per phrase in one run. One Google page takes 20–70 seconds, so keep maxImages small when you need a fast answer.

Troubleshooting

Fewer images than maxImages

Symptom: a phrase has fewer rows than you asked for, status exhausted. Likely reason: Google shows only about 300 images per phrase, fewer with strict filters or rare phrases. Action: remove a filter, or add more phrases with different words.

A phrase has no rows

Symptom: the phrase is not_found in RUN_SUMMARY. Likely reason: Google Images has no results for this phrase with these filters. Quotes and operators like site: make results narrower. Action: check the same phrase in Google Images in a browser, remove quotes or filters.

The run failed with "Google could not be read"

Symptom: run status error or partial with SOURCE_ERROR, phrases with error. Likely reason: Google did not answer after 3 tries. Rows saved before the error stay in the Dataset and are charged; failed pages are not charged. Action: run the failed phrases again later.

The run stopped as partial

Symptom: message "stopped at your spending limit" or "stopped at the run timeout". Action: raise the maximum cost per run, or the timeout, or split phrases into several runs.

FAQ

Do I need a Google account or an API key?

No. The Actor reads public Google Images results. No login and no key.

Are the images free to use?

Not always. The Actor gives links and does not give a licence. The usageRights filter is Google's label. Check the rights on the source page.

Why do two runs give different images?

Google changes results over time and between requests. In our tests the same phrase had about 10% different images between runs. Titles can also change with the language Google picks.

Can I get more than 300 images for one phrase?

No, Google does not show more. Use several similar phrases; repeated images are removed only inside one phrase, so check imageUrl across phrases yourself.

Support

For a bug, question or feature request, use the Issues tab of this Actor. Include the run ID, your input and the message from the run. Never send tokens or passwords.