Start with what you know.
If you have a question but no URLs, start with search. If you already know the page you want, start with extraction. If you have a list of pages or want to follow links within one site, create a collection. Choosing the right starting point keeps your workflow understandable and avoids unnecessary requests.
In pageflock, web search returns result titles, links and snippets — and, where Google surfaces them, a knowledge panel, a direct answer, related questions and a local business pack with phone, address and website — through an independently configured Serper connection. A result is a lead to investigate; the snippet is not the full page, and its wording can differ from the current source.
Read a page, then decide what to keep.
The pageflock Extraction Engine accepts a public URL and returns its title, readable text, links and metadata. It reads static HTML, with optional JavaScript rendering when a page needs it; login-only pages are outside this workflow. Dedicated extractors go further on a known page: contacts, product data with images and ratings, YouTube metadata with the description and stats, and a company profile that reads a whole company's public pages in one call.
Keep the source URL beside anything you save. Check the actual text before treating it as input to a report or model. For a search-led workflow, inspect the results first and extract only the relevant URLs.
Move to a collection when one page is not enough.
Use a batch when you already have your URLs. Use a same-origin crawl when the useful pages are connected within one website. A collection runs in the background and records each page's result. Your browser does not need to stay open.
Download the completed results as JSON or CSV within seven days. A successful search costs five credits, and each successful extracted page costs one. One search followed by three successful extractions therefore uses eight credits. Failed or blocked extraction is refunded.