How to own a new category keyword before it has volume

Because the payoff moved from the ranking to the citation, and citations are assigned before volume shows up. When an AI answer appears on a query, most of the clicks that used to spread across ten results concentrate on the few sources the answer names — and everyone else gets almost nothing. Pew Research Center measured the collapse: with an AI summary present, users clicked a traditional result on 8% of searches versus 15% without one, and clicked a source cited inside the summary on about 1% of visits.

The flip side is what makes the real estate valuable: being the cited source protects you from the collapse. Seer Interactive’s study of 3,119 informational queries found organic click-through fell 61% where AI Overviews appear — but brands actually cited in those answers earned 35% higher organic CTR than uncited brands on the same queries. The citation is the last position that still pays. On a new term, it is also the only position, and it is briefly free.

Pick the phrase a stranger would use to describe the behavior or problem — not the cleverest name, the most repeatable one. Three filters, in order:

• It describes, not brands. “Read to unlock” works because it says what happens. A term that embeds your product name caps its own adoption — rivals will never use it, and rival adoption is what grows the citation prize.
• It survives being spoken aloud. If a customer wouldn’t say it on a podcast unprompted, engines will never see it corroborated in the wild.
• It is checkably unclaimed. Ask the four engines what the term means. Hedged answers and borrowed vocabulary mean open; a consistent answer citing one source means you are displacing, not claiming — a different, more expensive play.

Then commit to one exact phrase, verbatim, everywhere. The signal you are building is co-occurrence — your name next to the term — and synonyms split it.

One page whose only job is to be the definition the engines lift. The structural spec is short: an explicit “X is …” definition in the first paragraph; question-shaped H2s matching what people actually ask; numbered steps or criteria wherever the content allows; and sections that answer standalone, because engines quote fragments and a fragment that depends on the previous section dies alone.

The evidence layer matters as much as the skeleton. The Princeton GEO study — the only controlled experiment on this so far — rewrote pages and measured visibility inside generated answers across 10,000 queries: adding quotations, statistics and cited sources lifted visibility 30–40% on the study’s position-adjusted word count metric, while keyword stuffing did roughly nothing. Engines are pattern-matching for pages that look like evidence, so the explainer should read like a well-sourced reference entry, not a landing page.

Here is the part of this genre we refuse to soften: you can execute the entire on-page spec above, perfectly, and never be cited. On-page structure decides which of your pages an engine lifts once it trusts you; it does not create the trust. The correlation data is blunt about where trust comes from — Ahrefs’ 75,000-brand analysis found branded web mentions correlate with AI visibility at 0.664, roughly triple backlinks at 0.218, and Seer Interactive’s research found much the same: strong Google-ranking and mention signals travel to LLM visibility, while backlink volume barely registers.

Even Google’s own documentation, which insists there are “no additional requirements” for appearing in AI features beyond standard SEO, points the same direction when read closely: the guidance is unique, expert-led content that satisfies the searcher — which, for a term you invented, means the wider web has to corroborate that your expertise is real. A definition only you assert is a press release. A definition the web repeats is a fact.

Not link building. The unit of value is the mention — your name and your term on a page you don’t control — and it is earned in places link builders ignore:

1. 01

   Get listed where your category’s directories live

   G2, Product Hunt, the App Store, an awesome-list on GitHub — whatever shelf your buyers browse. The listing text should use your category term verbatim, because directory rows are exactly the kind of corroborating page engines retrieve.
2. 02

   Answer the question where it is actually asked

   Reddit threads, Hacker News, niche forums, industry Slacks that leak to the web. Not astroturf — genuinely useful answers that happen to use the term. One honest forum answer that defines your category is worth more than ten pages on your own domain saying the same thing.
3. 03

   Give the term away

   Encourage competitors, reviewers and newsletter writers to use it. Every page that adopts your vocabulary — even to argue with you — corroborates that the category exists and that your definition is the reference.
4. 04

   Put the term in your unchangeable surfaces

   Product name suffixes, store listings, social bios, conference talk titles. Surfaces that third parties mirror and quote produce mention co-occurrence you never have to maintain.

None of this is fast, which is the honest reason the first-cited window is winnable at all: incumbents-in-waiting won’t do unglamorous seeding for a keyword their dashboard values at zero. The Ahrefs data even puts YouTube mentions at the top of the correlation table (0.737) — the single strongest brand-visibility factor they measured sits entirely off your website.

Not from your analytics — answer-engine visibility happens on other people’s screens, in conversations your tracking never sees. The only direct check is the obvious, tedious one: ask the engines your buyers’ questions, repeatedly, and count who gets named and whose pages get cited. Once is not enough; cited sources churn hard — AirOps puts it at roughly half of cited pages changing every month — so a single spot-check is an anecdote, and a lead you measured in March may be gone in May.

That measurement loop is the fifth step of the playbook, and it is the part we built a product for: a repeatable, sampled measurement of who the engines cite in your category, with a frozen prompt set so the re-scan is a real before/after. The PageLock teardown shows why it matters — the one step of that play nobody can observe from outside is whether it worked.