Owned vs. earned: how we bucket the sources AI actually cites

Knowing you were cited is step one. Knowing which URL the answer engine trusted — and whether it was yours — is what tells you what to publish next.

5 min readAnovox team

Where the source list comes from

Only some engines return a structured list of citations alongside their answer. For the rest, the reliable signal is the answer text itself: we extract every URL that appears in the stored response and normalize it down to a bare domain (no protocol, no www, no path). Each domain is counted once per appearance, and we keep the order in which it first showed up.

Owned, earned, and the domain you told us about

Your workspace has one domain — the site you actually control. Every cited source that matches it is 'owned.' Everything else is 'earned': review sites, comparison articles, Reddit threads, a competitor's blog. The split is reported as raw counts and as a percentage of all citations in the period, so you can see at a glance whether the AI is quoting you directly or quoting people who talk about you.

Attributing a source to a win or a loss

For every run, we already know whether your brand was cited and whether a competitor was. Each source domain that appeared in that run inherits both flags. Aggregated across runs, that gives three useful lists: the sources that appear when you win, the sources that appear when a competitor wins, and — most actionable — the sources that only ever show up alongside competitor wins and never alongside yours.

Why 'competitor favorites' is the list to read first

A domain that consistently helps a rival get cited and has never helped you is the clearest signal of where to earn a mention. It's usually a third-party site: a category roundup, a review aggregator, a community thread. Getting listed there is often faster than out-writing the competitor on your own site — which is why this list links straight into Agent Studio to draft the outreach or the comparison page.

What we deliberately don't do

We don't infer sources that weren't in the response. If an engine gave an answer with no URLs and no citation list, that run contributes nothing to this module. An empty state that says 'no sources detected yet' is more honest than a pie chart built on guesses.

See this on your own brand

Run the free check — same scoring the dashboard uses, no account needed.

Run free check

Related posts

How it works

How Aim ranks your gaps: severity × priority × recency × demand

6 min read
How-to

Your robots.txt Is Your GEO Front Door

4 min read