The crawl exports every address the old site answers, including the ones nobody remembers: old blog posts, PDFs, pages with query strings. Each is mapped to the page on the new site that answers the same question, one to one wherever one exists; a page with no successor is mapped to its nearest parent, never to the home page in bulk. The map is written as 301 redirects, tested on the preview against the full list, and shipped with the site.
After launch, Search Console is read for crawl errors and for the pages that earned impressions before, and anything that dropped is traced to its redirect. Titles, headings and schema are written for the new pages at build time, as on every site: how a new website gets found.