We Measured Our Own AI Citation Share Across Three Engines. Here's What Broke

We ran 23 seed queries relevant to our own positioning against three AI answer engines and tracked how often each one cited kompella.io. Perplexity cited us in 15 of 23 queries. Claude cited us in 11 of 23. Google's AI Overviews cited us in 10 of 23. Same queries, same site, three very different outcomes. If you're treating "AI visibility" as one number, this is why that's a mistake.

The numbers, engine by engine

We're publishing our own citation-share data because most of what circulates about AI search visibility is vendor marketing with no methodology attached. Here's ours, plain.

• Perplexity: 15 of 23 seed queries cited kompella.io, a 65.2% citation share.
• Claude: 11 of 23 seed queries cited kompella.io, a 47.8% citation share.
• Google AI Overviews: 10 of 23 seed queries cited kompella.io, a 47.6% citation share.

First, the spread between engines is real and not small. Perplexity cited us in roughly two out of three queries. The other two engines cited us in under half. If you're optimizing for "AI visibility" as a single undifferentiated target, you're optimizing for nothing. These are three different retrieval and ranking systems with different citation behavior, and a strategy tuned for one won't automatically transfer to the others.

Second, Claude and Google AIO landed within a fraction of a point of each other (47.8% versus 47.6%) despite being built entirely differently. That's either coincidence at this sample size or a sign that both are pulling from a similar layer of the web's authority signals. With 23 queries per engine we can't tell you which. We can tell you the numbers matched almost exactly, and that's worth watching over more runs before anyone builds a theory on it.

Third, this whole exercise runs on a 23-query seed set. That's a starting instrument, not a definitive index of "our AI visibility." Treat every percentage above as what it is: the citation rate on this specific query set, on this specific date, for these three engines. It'll move.

The citation landscape underneath those numbers

Citation share only tells you whether we showed up. It says nothing about who else did, or how competitive the field is. So we also looked at the full set of domains cited across the same query run.

Across the 23 queries, 292 distinct domains got cited at least once by at least one engine. That's the real competitive set, not a handful of named competitors, a genuinely long tail of practitioner sites, aggregators, and platforms.

The most-cited single domain in the set showed up in 46 citations across 17 of the 23 queries, with an average cited rank of 9.65. That's a domain the engines return to constantly and rank fairly low when they do, worth noting because "gets cited a lot" and "gets cited prominently" are two different wins and this domain has one, not both.

Two structural patterns matter more than any single domain's rank:

1. The list is diluted by non-practitioners. Of the ten most-cited domains overall, three are general platforms or job boards rather than firms actually doing the work being searched for. If you're benchmarking yourself against "the top 10 cited sites" for your category, check whether you're actually benchmarking against a jobs board first. 2. The distribution is long-tailed, hard. Most of the 292 cited domains appear only once or twice across the entire 23-query run. A small number of domains get cited repeatedly across many queries; the overwhelming majority get pulled in for one specific question and never again. That means consistent, repeated citation across a query set, not a single high-ranking hit, is the harder and more valuable thing to earn.

Why this matters more than your click-through rate

Here's the uncomfortable second half of this data. In the same measurement window, our Search Console numbers looked like this: 32,727 total impressions, 120 clicks, a 0.37% overall click-through rate. Restrict that to human traffic only (stripping out bot and agent crawl activity) and it's 15,892 impressions, 53 clicks, 0.33% CTR.

A 0.33% human CTR is not a typo. It's what happens when your content gets pulled into AI answers, summarized, and cited without a human ever needing to click through to your page. The impression still counts. The click mostly doesn't happen. Of the impressions in that window, 15.5% were classified as bot or agent traffic rather than human, which is its own signal that machine consumption of your pages is no longer a rounding error.

If your reporting stack still treats organic clicks as the primary success metric, you are structurally blind to the channel that's actually consuming your content now. Citation share, not CTR, is the metric that reflects whether AI engines trust your site enough to surface it in an answer.

What we're doing with this

We're not chasing a single "AI SEO score." We're tracking citation share per engine, per query, over time, because the engines behave differently and averaging them into one number erases the information that matters. A site that's strong on Perplexity and weak on Google AIO needs a different fix than a site that's weak everywhere, and you can't tell the difference without measuring per-engine.

The practical takeaway if you're running this for your own site: pull your own seed query set (the questions your buyers actually ask), run it against each engine you care about, and count citations per engine separately. Don't average. The average hides the exact thing you need to act on.

Need help thinking this through? Book a 30-min call, no pitch.

FAQ

Need help thinking this through? Book a 30-min call — no pitch. Book a Free 30-Min Strategy Call →

Data and method

Every figure above is read directly from a measurement artifact committed to the kompella.io repository. None of them is an estimate, an industry average, or a round number chosen to sound credible. Snapshot: 2026-08-11 05:02 UTC; 2026-08-10 04:30 UTC.