Backlinks were a volume game: more good links, more authority, better rankings. Citations are not. In every category we track, a small, stable set of domains supplies most of the sources a model leans on, and the set barely moves week to week.
The concentration problem #
In the payments category we tracked, 12 domains supplied roughly half of all citations across 900 answers. In developer tools it was nine. In HR software, fifteen. The long tail exists, but it is decorative.
That changes the strategy. You are not trying to accumulate citations. You are trying to get into a shortlist that already exists.
Three kinds of citation #
Not every citation is worth the same, and the split is worth tracking separately:
- Owned: your own domain. Cheap to influence, weakest signal.
- Earned: third-party reviews, comparisons, press. Slow, strongest signal.
- Rival: a competitor’s page cited in an answer about you. Worth an alert.
A healthy mix in our data ran roughly 30 / 45 / 25. Teams heavy on owned citations tended to be quoted for what they do but not recommended.
The fastest wins came from correcting outdated third-party comparison pages, not from publishing anything new.
Where to start #
Pull the domains cited across your tracked prompts and sort by frequency. Then, for the top ten:
- Check whether your entry is present and current.
- Check what claim it makes about you.
- Check whether a rival’s entry is more specific than yours.
Most teams find at least two pages in that top ten with stale pricing or a missing product line.
Related reading: how position beats presence and writing pages engines can quote.