The reverse image search API takes a publicly reachable image URL and returns where that image appears across the web — one row per visual match, with the match title, the page it lives on, the source site and its domain, plus image and thumbnail URLs. It is Google Lens' reverse search, delivered as data instead of a results page.
It runs in two modes. By default it returns visual matches — pages showing the same or a closely similar image. Set exact_matches and it reads the 'pages that use this exact image' tab instead, which adds the publish date and pixel size per entry — the mode brand and rights teams use to find unlicensed reuse of a specific asset. Each run returns up to 40 matches over the request tier, no browser, and the image you pass only needs to be reachable at a public http(s) URL.
$0.002 per delivered match, up to 40 matchs per run. Nothing delivered means nothing charged.
You pass an image_url that Google can fetch, optionally set exact_matches, country/lang and a max_results up to 40. The collector runs the reverse search through an in-region exit and flattens each match into a row.
Two behaviours matter:
exact_match stamped false. With exact_matches set, it reads the exact-image tab and each row gains date and size, with exact_match true — the split you want for rights work versus discovery.image_url is required and must be a publicly reachable http(s) image. exact_matches switches from 'similar images' to 'pages using this exact image' (adding date and pixel size per row). country/lang set the exit and locale; max_results caps delivery at 40 matches.
Each match is one row: title, link to the hosting page, source site name and its registrable domain, plus image and thumbnail URLs. In exact-match runs date and size are populated and exact_match is true; in the default visual mode those two are null and exact_match is false, so you can tell the modes apart in a merged table.
$0.002 per delivered match ($2 per 1,000). Nothing delivered means nothing charged, and the $2 monthly free credit covers roughly 1,000 matchs here. Volume tiers take up to 30% off.
$0.002 per delivered match, $2 per 1,000. One image returns up to 40 matches, so a lookup is at most about eight cents. A brand-protection sweep of a few hundred hero images is single-digit dollars, and you pay per match returned — an image with few matches costs proportionally less.
The $2 monthly allowance (no card) covers about 1,000 delivered matches, and volume tiers take up to 30% off. GET /v1/scraper/collectors reports your key's live price.
Google's own programmatic reverse-image capability ships as Cloud Vision web detection, not as a Lens API — there is no public 'Google Lens API'. Cloud Vision can return pages with matching images, but it is a different product to stand up: a Google Cloud project, a billing account, an enabled API and its own quotas and per-image pricing, and its web-detection response is shaped around entities and confidence rather than the Lens results a person would see.
This collector returns the Lens surface directly — the pages, sources and domains from the visual and exact-match tabs — under your existing QuantumProxies key, with an in-region exit and a flat per-match price. If you specifically need Vision's entity labels or safe-search signals, that is the first-party tool; if you need 'where does this image appear', this is the shorter path.
Driving Lens by hand is deceptively involved:
date/size that only exact mode exposes.Want to keep your own pipeline and just run it on clean exits? That network is our residential proxies.
Run your product or campaign images with exact_matches set and get every page reusing the exact file, each with a publish date and pixel size — the evidence a takedown needs.
Reverse-search a catalogue photo to find marketplaces and sites showing the same image, a fast way to surface unauthorised resellers using your own product shots.
Given an image, find the earliest and most authoritative pages hosting it via source and domain — useful for verification, licensing and de-duplication work.
One run reverse-searches one image_url and returns up to 40 matches. The image must be publicly fetchable by Google — private, authenticated or localhost URLs will not work. Visual mode returns similar-image matches; exact mode returns pages using the identical file with date and size. Fields Google omits come back null, never guessed. Runs parallelise up to your plan's rate limit — 60 a minute on pay-as-you-go, up to 1,200 on the top tier.
Legally: the pages returned are public, and reading public data is broadly lawful, but Google's terms restrict automated access, image copyright is a separate matter from discovery, and rules vary by country. Background, not legal advice.
$2 of usage a month with no card, which at $0.002 per match is about 1,000 delivered matches. There is no public Google Lens API to compare against; Google's nearest first-party option is Cloud Vision web detection, billed per image on a Cloud account.
No. You authenticate with your QuantumProxies API key only — no Google Cloud project, no billing account and no Vision API enablement. You pass an image URL and get matches back.
Default mode returns visual matches — similar images across the web. Setting exact_matches reads the 'pages using this exact image' tab instead, adding a publish date and pixel size to each row, which is the mode used for rights and takedown work.
Not directly — the collector takes a public image URL, because that is what Google fetches to run the search. Host the image at a reachable http(s) address first, then pass that URL as image_url.