Measuring AI citations: how to track whether engines cite your brand.

Four things, in order of how quickly they tell you something useful:

• Coverage: out of the engines you check, how many cite your brand at all? This is the are-we-on-the-map number.
• Density: across repeated trials, how often does your brand appear? Coverage says whether, density says how often.
• Share of citations: for the prompts where you are absent, which competitors get named instead? This is the versus-whom number, and it is the one that points at the work.
• Accuracy: when an engine does cite you, is it describing your product correctly, or confusing you with someone else, or naming an old version?

AI answers drift, sometimes week to week. They personalize to a logged-in account's history. They vary by region. A single prompt run once on a single engine is N=1, and N=1 cannot tell a real pattern from a fluke.

A measurement repeats the same prompts across engines and across trials, so what you record is a pattern rather than a coincidence. That is the difference between "I checked and we were not there" and "we are cited by one engine of four, in roughly a third of trials, and never by name in the others."

There is a manual version and a rigorous version, and both have a place.

The manual version is our 10-minute AI search audit: five buyer prompts across four engines, scored by hand. It is enough to answer the screening question, are we invisible. The step-by-step of running that check, including the mentioned-versus-cited distinction, is in how to check if AI search mentions your brand.

The rigorous version is what our SXO Audit runs: 25 buyer prompts across six engines with three trials each, region-aware, repeated so the result is reproducible. The same method produced our Invisible 10 study, where 600 responses across four engines returned zero citations of ten funded vendors.

If you track one number, track share of citations on your target queries, over time. Absolute counts move whenever an engine changes its retrieval or its model, so a raw count up or down tells you little on its own. Your trend, and your position relative to the competitors the engines name instead of you, are the signals that survive those changes.

Can I measure AI citations for free?

Yes. A manual ten-minute audit across the four main engines gives you a yes-or-no read at no cost, and our Free Snapshot checks one buyer prompt against your domain in exchange for your email. Paid tools add rigor, not the basic answer.

How often do AI answers change?

Often enough that a one-time check ages quickly. Engines update their retrieval and their models on no fixed public schedule, and we have seen answers to the same prompt shift within a week. Measure on a cadence rather than once.

What counts as a good citation rate?

There is no universal number. The honest benchmark is your own trend over time plus your share of citations versus the specific competitors the engines name instead of you. A rate that is climbing against named rivals matters more than any absolute figure.

Start with the manual audit, or have us run the reproducible version across six engines.

Try the Free Snapshot   See the SXO Audit

By the Web Cited Editorial Research Team. Last updated 31 May 2026.