Keyword research for a brand new website
Every keyword guide assumes existing traffic to analyse. You have none, and the advice does not apply.
You are not choosing keywords, you are choosing which demand to intercept
You cannot afford to create demand. You can afford to be there when it arrives. Start from problems people already have and already type, not from what you wish they searched for.
Count the sub-problems in your category
Almost every category splits into far more distinct jobs than it first appears. One product might serve sixty separate queries, each a page. That count, done honestly on a whiteboard in an afternoon, is a better plan than any volume export.
Pick the ones you can win now, and revisit in a quarter
At zero authority, target the queries whose page-one results are weakest. Your ceiling rises as links accumulate, and the head terms that are impossible today may be reachable in a year. They will still be there.
Start from what you can prove, not what you want to rank for
A new site has no authority, which means the only terms available are the ones nobody with authority has bothered to cover well. Those are found by reading results pages, not by sorting a spreadsheet by volume. If page one is three forums and a Reddit thread, that is an opening. If page one is four sites at DR 80, that is a term for year three.
The first ten pages decide the next hundred
Whatever you rank for first shapes what Google is willing to rank you for next, because the internal links, the crawl attention and the topical association all compound from there. Pick the first ten so they point at the same thing. This site's first pages were tools, and every later win was adjacent to them. That was not planned, but it was decisive.
Measured, not asserted
This site began at zero on 25 December 2025 and reached 100,542 users by 5 August 2026 with $0 spent on advertising, on 160 pages built this way.
Free tool for this: SERP Audit. No account, nothing uploaded.
Where this goes deeper
Every number on this page comes from one complete dataset: one product taken from zero to 100K+ users on search alone, with nothing spent on advertising. The full argument is Chapter 1 and Chapter 6 of the book. Five chapters are free to read.