Search Google

Run a Google search and return the page's organic result cards (title, url, site, byline, meta, snippet); the /url?q= redirect wrapper is stripped from urls. Runs with your injected Google cookies (loggedIn:true), so results are personalized to your account/IP. QUERY — full Google operators work: site:, quoted "exact phrases", OR, -exclude, intitle:. Quoting phrases narrows to a small, high-precision set (with an occasional unrelated tail); dropping quotes broadens recall, and the extra hits tend to be adjacent near-misses rather than random noise. Use site: to scope a domain/path (e.g. site:www.linkedin.com/in). PAGINATION — one call returns ONE results page: EVERY organic card on it (~10), no cap. To reach further results use `page` (0-based, maps to Google's &start=page*10), like the SERP pager. Pages replay deterministically (same account/IP) and are NEAR-disjoint — adjacent pages occasionally share ~1 boundary result — so you can fetch different pages CONCURRENTLY and union them; always dedupe by canonical url. (Google retired the &num depth trick — n>10 never deepened — so there is no per-call count; page is the only lever.) DETERMINISM — narrow queries reproduce exactly (same members AND order) across replays on one account/IP; very broad / large-result queries may reorder between runs. MORE COVERAGE — page the query (concurrent page fetches → union) and/or partition the query space: vary the site: path, phrases/OR-terms, or `freshness` windows (which return disjoint, correctly-dated sets). Always dedupe by url.