The citation gap: finding the sites that already cover your competitor but have never once cited you

Share of voice tells you that you are behind. It does not tell you where to go

'A competitor has a higher share of voice than you' is true and almost useless as an action item on its own. It names a symptom, not a cause, and definitely not a next step. The useful follow-up question is narrower and far more actionable: which specific sources, the ones AI engines actually lean on when they answer questions about this market, already cover that competitor and have never once mentioned us?

That is a different, answerable question, and the answer is a short, concrete list: named domains, not an abstract percentage gap.

How a citation gap is actually computed

Every answer an engine returns with web search enabled can come with citations, the sources the engine itself displayed or referenced. Those citations are already stored against every run in an audit, classified by domain, and already used to compute ordinary source-intelligence metrics such as citation rate and which domains are cited most.

A citation gap reuses exactly that stored data, for a specific tracked competitor. For each competitor, it finds every run in the audit where that competitor was actually named, then checks which domains were cited specifically in those runs. Of those domains, the ones that were never once cited in a run naming the brand doing the audit, and are not that brand's own domain, are the citation gap for that competitor.

This means no new data collection happens to compute it. It is a different lens on citations the audit already captured, which is also why it costs nothing extra to compute and updates automatically every time a new audit runs.

The head-to-head number next to it: confident, possible, or even

Alongside the citation gap sits a second, related number: how a competitor's own visibility rate compares to the brand's own visibility rate, in the same audit. This uses the identical statistical method already used to decide whether a change between two audits is real or just noise, a Newcombe interval on the difference between two rates. Applied here, it answers whether a competitor's lead (or deficit) is large enough, relative to the sample size behind it, to call it a confident or merely possible lead, rather than quietly treating any numeric difference as if it were already proven.

A competitor's rate 8 points above yours, on a four-run sample, is not the same claim as the same 8-point gap on a 60-run sample, and the labelling says so directly: confident, possible, or about even, never a bare, unqualified percentage difference standing in for a verdict it cannot actually support.

Turning a gap list into an actual plan

• Start with domains that cite a competitor across several different prompts, not just one, since that is a stronger, more durable pattern than a single mention.
• Check what kind of source each gap domain actually is: a review site, an industry directory, a comparison article, a forum thread. Each needs a different kind of outreach or content response.
• Treat a citation gap as a prioritised list to work through, not a one-time audit finding. It is recomputed every time a new audit runs, so progress is visible the same way the gap itself was found.