From a photo to the exact retailer URL

Insurance-claim software (and plenty of other pipelines) detects a product in a photo, then needs the exact retailer product page to price it — the page a human would actually buy it on. The obvious path, a Google Shopping search, hands back a tracking redirect (google.com/…&udm=28), not the merchant's page. And the "direct link" fields the product APIs expose are inconsistent and often missing.

The robust fix is to stop trying to repair that redirect and instead pull the URL from a source that returns real retailer pages — here, Google Lens straight from the cropped image, in a single call:

Pick a sample above and hit Resolve — the server runs the lookup live, then ranks the matches by retailer domain, price, and exact-match confidence, and falls back gracefully when nothing's confident.

A few things worth noting:

• One call, from the crop. No search → scrape → resolve chain; the match, the price, and the real product URL come back together.
• Cheaper variant for production. Skip the image and run an organic web search on the brand + name you already have, scoped to the detected vendor's domain ("<brand> <name>" site:<vendor>). It costs pennies per thousand and you already hold the price from the existing shopping call.
• Never trust one field. Real items return marketplaces, wrong variants, and dead matches — so the pipeline ranks candidates and degrades to "needs review" rather than shipping a confident-but-wrong link.