Schwartz ran search at a well-known survey company and writes against the industry he comes from: agencies that hand over keyword lists and technical audits nobody executes. His central idea is that this is not a marketing channel bolted on at the end but a product decision: you build pages that do a job for whoever lands on them, and the search engine rewards them as a consequence.
What I took from it
The order of the questions. Before writing a page, the question is not what I want to say but whether anyone types that, in those words, when they have the problem. A flawless page about something nobody searches for is good work that produces nothing, and no ranking report will ever show you that difference.
The split between pages that inform and pages that do something. A table that settles a specific question, a checker, a calculator: that is product. It holds up on its own, it does not age like an opinion, and whoever arrives knows within three seconds whether it helps. Since then, when I think of a new page, I look first for the task it could finish right there.
And an expectation correction I use with clients: position is an intermediate indicator, not a result. Moving from ninth to third is internal news; the result is someone doing something once they arrive.
What I left out
The scale. Schwartz writes for companies with product and engineering teams able to generate tens of thousands of pages out of a database, and a good part of the book goes there. Someone working alone cannot execute that and does not need it. The book is also from 2021: it never met the search engine that answers without anyone visiting the site, and that change hits precisely the purely informational pages he already advised against.