Google Lens Visual Search API — Image Matches
Pricing
from $12.50 / 1,000 completed visual-search requests
Google Lens Visual Search API — Image Matches
Find Google visual-search matches for public image URLs. Returns Google-associated query context, organic source pages, image matches, refinement chips, and source provenance through a managed DataForSEO backend.
Pricing
from $12.50 / 1,000 completed visual-search requests
Rating
0.0
(0)
Developer
Khadin Akbar
Maintained by CommunityActor stats
0
Bookmarked
5
Total users
1
Monthly active users
an hour ago
Last modified
Categories
Features
Share
What is Google Lens Visual Search API?
Give it image links and this Google Lens Visual Search API returns a dataset of where each image appears on the web: up to 60 page links, up to 10 titled results with a short description, and Google's best-guess search phrase for every image. It is built for product researchers, brand-protection teams, marketplace sellers, and AI agents that need to trace an image back to its sources.
- Up to 100 images per run, checked in parallel. Free plan runs check up to 10 images.
- Pay only for images that find matches. An image with no match is saved as a free row.
- Plain results. One row per image with a status, a match count and a one-sentence summary.
Best fit
Choose it when an image is your starting point and the next decision depends on where that image appears online: finding the original source of a product photo, spotting resellers or copies of your catalog images, checking where artwork or a logo is reused, or giving an AI agent a quick evidence list for an image.
Example workflow: a marketplace seller pastes 40 supplier product photos, runs the Actor, and filters the rows with status of matches. The imageMatches links show which shops already sell the same item, and the domain column shows the sites that show up most. The seller uses that list to compare prices and pick suppliers.
What you get for each image
| Field | What it tells you |
|---|---|
status | matches when Google found something (charged), no_matches when it found nothing (free). |
matchCount | Page links plus titled results found for this image. |
agentSummary | One sentence summary, handy for people and AI agents. |
imageMatches | Up to 60 real page links where the same or a similar image appears: rank, url, domain. |
organicMatches | Up to 10 titled results: title, description, websiteName, the source domain, and a Google redirect url. |
googleAssociatedQuery | Google's best-guess search phrase for the image, when it has one. |
refinementChips | Suggested ways to narrow the search, when Google shows them. |
checkUrl | A Google address you can open to check the search yourself. |
imageUrl, locationCode, languageCode, retrievedAt, providerTaskId | Which image, which market and language, when it was collected, and a task ID for support. |
Example row
Sample of a real row, shortened to three page links and two titled results.
{"imageUrl": "https://upload.wikimedia.org/wikipedia/commons/4/45/A_small_cup_of_coffee.JPG","status": "matches","matchCount": 69,"agentSummary": "Found 60 page links and 9 titled results. Most frequent sites: facebook.com, world.nol.com, instagram.com.","organicMatches": [{"rank": 1,"title": "Coffee - Wikipedia","urlType": "google_redirect","domain": "hif.wikipedia.org","websiteName": "Wikipedia","isImage": true,"isVideo": false}],"imageMatches": [{ "rank": 1, "url": "https://nihonkoku-shoukan.fandom.com/wiki/Coffee", "domain": "nihonkoku-shoukan.fandom.com" },{ "rank": 2, "url": "https://en.wikivoyage.org/wiki/Coffee", "domain": "en.wikivoyage.org" },{ "rank": 3, "url": "https://sco.wikipedia.org/wiki/Coffee", "domain": "sco.wikipedia.org" }]}
The url on a titled result is a Google redirect link (urlType says so). The domain next to it is the site Google shows for that result. The imageMatches links are direct page addresses.
How to use it
- Paste one or more public image links (a direct link to the image file, not a web page).
- Pick how many matches you want per image (up to 60) and, if needed, a market and language.
- Start the run. Results arrive in the dataset as each image finishes.
{"imageUrls": ["https://upload.wikimedia.org/wikipedia/commons/4/45/A_small_cup_of_coffee.JPG"],"maxResults": 20,"locationCode": 2840,"languageCode": "en"}
locationCode is a numeric Google location ID; 2840 means the United States. languageCode is a two-letter language such as en, de or fr.
Call it from code
curl -X POST 'https://api.apify.com/v2/acts/khadinakbar~google-lens-visual-search/runs?token=YOUR_APIFY_TOKEN' \-H 'Content-Type: application/json' \-d '{"imageUrls": ["https://upload.wikimedia.org/wikipedia/commons/4/45/A_small_cup_of_coffee.JPG"], "maxResults": 20}'
Read the rows from the run's default dataset. The OUTPUT record in the key-value store holds a short summary: how many images were charged, how many had no match, and any warnings.
Pricing and cost control
This Actor uses Pay per event (PPE): you pay per image that finds at least one match, plus a very small start fee per run. Apify platform usage is included in the event price. Images with no match and images that fail are free. The live Pricing tab is the source of truth for the current price.
| Run | Images charged |
|---|---|
| 1 image with matches | 1 |
| 10 images, 8 with matches | 8 |
| 100 images, 90 with matches | 90 |
The run also respects your Apify maximum-charge limit. If the limit would not cover every image, the Actor checks only as many images as the limit can pay for and says so in the run summary, so nothing is searched that cannot be paid for. Free plan runs are capped at 10 images and may stop early when the shared free allowance for the day is used.
Best results
Start with a stable, direct link to a clear image, such as a product hero shot or a catalog photo with one subject. Choose the market (locationCode) and language that match where you are researching, keep maxResults at 20 for a quick scan, and raise it to 60 when you want every link Google shows. Open checkUrl to see the same search on Google.
Use it from an AI agent
Run Google Lens Visual Search API on these image links. When it finishes, read the dataset, keep rows with
statusofmatches, list the topdomainvalues fromimageMatches, and report how many images were charged from theOUTPUTrecord.
The run is small to describe and the output states its own scope: each row has status, matchCount and agentSummary, and OUTPUT lists charged images, free images and warnings.
What's new
- The result rows are now full. Earlier versions returned a single empty-looking row; every image now returns its real page links (up to 60).
- Up to 100 images per run, checked in parallel (it was 10).
- Images with no match are free.
- A one-sentence
agentSummaryand astatuson every row. - Free plan runs are capped at 10 images.
Good to know
- Use a direct link to a public image file; web page links, private-network addresses and local files are rejected with a clear message.
- Built for image links; to read the text inside an image, pair it with Google Lens OCR, and use a hosted copy of any local file.
- Google can return nothing for an unusual image; that row is saved as
no_matchesand costs nothing. - Speed: a single image takes about 20 to 45 seconds. Large runs check about 20 images at a time, so 100 images take around four to five minutes. If a run is close to its time limit it stops starting new images, says how many were left, and charges nothing for them; send those again in a new run.
- Results reflect what Google showed at
retrievedAtand can change over time.
FAQ
Why do some titled results show a Google link instead of the site address?
Google hides the final address behind its own redirect for those results. The domain field shows the site Google displays for them, and imageMatches gives direct page addresses.
How many links will I get per image?
Usually up to about 60 page links and up to 10 titled results. The number depends on how widely the image appears; maxResults caps it.
Am I charged when nothing is found? No. Only images with at least one match are charged.
Why did my run check fewer images than I sent? Free plan runs check up to 10 images, and any run checks only as many images as its maximum-charge limit can pay for. The run summary says which applied.
Can I use it from an AI agent?
Yes. The input is small, every row carries status, matchCount and an agentSummary, and the run's OUTPUT record states what was charged.
Related Actors
Next step in a workflow: Google Lens OCR reads the text inside an image, so you can combine where an image appears with what it says.
Builder's note
I found that the useful part of a Google visual search is the long list of page links tucked under the "Visual matches" block, and that earlier versions of this Actor read only the block's heading. I learned to read those nested links directly, to keep every search to one result page so each image stays cheap, and to charge only when an image really finds something.
Responsible use
This Actor is not affiliated with, endorsed by, or sponsored by Google. Use it for products, logos, artwork and similar images you are allowed to analyse. It is not meant for identifying private individuals; do not use it to find or track a person. Follow applicable laws, copyright rules and the terms of the sites you visit, and confirm ownership and licensing before acting on a match.

