Google now hides every search result URL behind /goto redirects, making SERP scraping O(n) requests.

Direct result links are replaced with google.com/goto?url= carrying a proprietary, non-base64 encoding that acts as an opaque reference to an index record. Recovering a destination needs a separate request to the /goto endpoint reading the Location header without following it, so harvesting one result page goes from a single parse to hundreds of sequential callbacks to Google. Consistent across logged-out and private-mode searches since late August 2026, following removal of &num=100 pagination and tighter BotGuard detection (autom.dev, 475 points on Hacker News). Whatever you built on SERP parsing just got a few hundred times more expensive per page.