A brand-new niche is keyword research on hard mode: no rankings to read, no analytics to mine, sometimes not even competitors to copy. Everything has to come from public sources.

Good news: the public sources are enough. Here is the zero-to-list pipeline, step by step.

Step 1: harvest vocabulary before generating anything

Generators need seeds, and in a niche you barely know, your own guesses are the worst seeds available. The audience already wrote better ones.

SourceVocabulary it yields
Reddit and niche forumsSlang, recurring problems, "what do you call the thing that..."
Amazon and store reviewsProduct language, complaints, use cases you never imagined
Docs and glossariesThe technical terms practitioners search
YouTube commentsSpoken phrasings and beginner confusion, verbatim

Twenty minutes of reading, one growing note of terms people actually use. Ten to fifteen seeds is the harvest target, and every one of them is pre-validated by existing usage.

Step 2: expand through autocomplete

Feed the harvested seeds to the generator on this site and let autocomplete do what it does: return every real phrasing typed around each seed, streamed live.

Autocomplete matters double in new niches because it updates faster than volume databases. It shows what people type now, including phrasings too fresh to have reported numbers.

Add the cursor and prefix tricks on your two strongest seeds, and the question sweeps for the informational layer. A few hundred candidates is normal from here.

Step 3: run the modifier matrix

Cross your seeds with the 120-modifier list, leaning on the beginner, problem and comparison groups. New niches are beginner-heavy by definition, and their comparison queries are usually unclaimed.

The matrix also drafts your site structure for free: modifier groups that produce many real phrases are content categories waiting to be named.

Step 4: validate, with adjusted expectations

Run the shortlist through the volume tools, then read the numbers differently than you would in a mature niche.

Volume reported demand is proven and everyone can see it: expect competition target the long-tail versions first In autocomplete, no volume people type it, databases have not caught up: demand newer than data the new-niche sweet spot, claim it before the numbers In neither maybe too early, maybe nobody wants it: this is a bet, not a keyword a few bets are fine, a site of bets is not
The middle column is where new-niche sites win: real typing, no reported numbers, no competition yet.

In mature niches, zero-volume means skip. In new niches, autocomplete presence with zero reported volume is often the earliest visible form of demand.

Step 5: pick first targets like an underdog

A new site cannot win head terms, and does not need to. Order the list as a ladder.

months 1-3: specific questions and long-tail months 3-9: comparisons and best-of lists later: category head terms winnable now, builds topical depth earns links and buyers reachable once the lower rungs rank
Each rung makes the next one reachable. Skipping rungs is how new sites publish into silence.

One hundred validated keywords, laddered, is a launch plan. Write the bottom rung first, and let three months of real ranking data replace this whole zero-data method with something better: your own numbers.

The one-line takeaway: harvest the audience's own vocabulary, expand it through autocomplete, matrix it with modifiers, and trust autocomplete presence over missing volume data. New niches are won on the long-tail rungs first.