Being found now means being cited

For twenty years the discovery problem had one shape. A buyer typed a query, a search engine returned ten links, and the work was to be one of them. Ranking was the whole game, and everything downstream, from keyword research to backlink building, was in service of a position on a page.

That shape is gone for a growing share of queries. Ask an assistant a real question and you don’t get ten links. You get an answer, with two or three citations, and most of the time nobody clicks anything.

Two different jobs

Indexing and citation are not the same problem, and treating them as one is where most discovery work now goes wrong.

Indexing rewards coverage, structure and technical health. It’s a crawler deciding whether your page is eligible. Citation rewards something else: whether a paragraph can be lifted out of your page and stand alone as a correct answer to a specific question. Those are different properties. A page can rank third for a term and never be quoted, because its answer is spread across six sections and buried under a brand story.

So the practical work splits. Keep the technical foundation clean, because a crawler that can’t parse you can’t cite you either. Then, separately, rewrite for extractability: one question per section, the answer in the first sentence, the specifics (numbers, names, dates) present rather than implied.

What extractable content looks like

The test is mechanical. Take a paragraph out of your page, with no surrounding context, and read it in isolation. If it still answers something and still sounds like it was written by someone who knows, it can be cited. If it says “as we mentioned above” or opens with a preamble, it can’t.

In practice that means:

• Headings phrased as the buyer’s actual question, not as a section label.
• The answer in the first sentence under that heading, not after a warm-up.
• Concrete detail over positioning language. Specifics are what make a passage quotable.
• Machine-readable structure: schema, clean headings, a summary file the assistant can parse.

The measurement problem

Rank tracking doesn’t measure any of this. You can watch a query every day and learn nothing about whether an assistant is citing you, because the citation isn’t in a rank position. You have to go and ask, repeatedly, from more than one assistant, and record what comes back.

That’s unglamorous work and almost nobody does it consistently. Which is exactly why it’s worth doing: a tracked set of twenty buyer questions, checked monthly across two or three assistants, produces a dataset no competitor has. Over a quarter you can see whether rewrites moved citation, which is a real feedback loop rather than a guess.

Rank position is a proxy for visibility. Citation is visibility.

What doesn’t change

Traffic still matters, and citations still produce some. What changes is the ratio. If a third of your category’s research happens inside an assistant, the pages that get cited collect the demand that used to spread across ten results, and the pages that merely rank collect the remainder.

The honest trade-off: optimising for citation usually means writing narrower, more specific pages. That can reduce the surface area that ranks for broad terms. You’re trading breadth for quotability, and the right answer depends on whether your buyer researches in a browser or a chat window. Most B2B buyers now do both.