Citation share is the cleanest scoreboard answer engines have produced yet: the percentage of AI answers across a query set that name your domain as a source, your share of voice in a world where the answer arrives pre-summarized. The number is real, and it is worth watching.
What the dashboards selling it leave out is simpler: you cannot move citation share directly. It behaves like a rank, a readout of inputs you set earlier rather than a dial you reach in and turn. You spent a decade learning this with search rankings: you set the inputs, and the position followed.
The input citation share reads is provenance, whether the load-bearing numbers on your page originate there or point one hop upstream to whoever you borrowed them from. We traced 1,469 SaaS blog citations to the end of their reference trail to see where that resolution lands, and only 15.0% reached a primary source.
The slot that wins citation share sits open on most pages. Almost nobody is counting whether they hold it.
What Citation Share Means in AI Search
Citation share is the percentage of AI answers across a fixed query set that cite your domain as a source, measured against every other cited domain. It is the answer-engine version of share of voice: your cited answers divided by all cited answers, times 100. It is a percentage, not a position.
As a formula:
citation share = (your cited answers / total cited answers) × 100
You sample it separately on each engine, because ChatGPT, Perplexity, and Google's AI answers each assemble sources differently, and you read it as an outcome of answer engine optimization rather than a lever inside it. That distinction decides everything that follows.
Citation Share vs Share of Voice
Share of voice measured your slice of attention across a channel: paid impressions, organic rankings, social mentions. Citation share measures the same slice one layer in, across the sources an answer engine credits when it composes a reply. Share of voice counted where you appeared. Citation share counts where you were believed enough to be named as the source.
The mechanics rhyme, and so does the mistake. Teams treated share of voice as a budget they could buy their way up, until the channels that mattered stopped selling impressions outright. Citation share inherits the same trap in a sharper form. There is no media buy for being the place a number originates. The slice you hold is decided before the engine ever assembles the answer, by whether the figure on your page traces back to you or to someone upstream.
Why Citation Share Behaves Like a Rank Not a Dial
Citation share behaves like a rank because it sits at the end of a process, not at the controls. You spent years learning this with search rankings. You never opened a console and set yourself to position three. You changed the inputs an algorithm reads, and the position settled where those inputs put it. The number on the dashboard was a report, and the work happened upstream of the report.
A citation-share dashboard is the same kind of report, aimed at a newer surface. It reads how often answers already cited you, and it cannot reach back and make the next answer cite you more. The vendors selling the dashboard rarely say this, because a thermometer is easier to sell than the harder claim underneath it: the percentage moves only when the thing it measures moves, and the thing it measures is provenance.
Provenance is the input. How do you check for it today?
Poll: How do you check whether AI search is citing you?
Four of these five answers watch an outcome. Only one checks an input.
A tracker, a referral chart, a manual spot-check all read the scoreboard after the engine has already chosen. The one method that touches what the scoreboard reads is checking whose numbers sit on your page in the first place. As readers weigh in above, the split between watching share and owning what produces it will come into focus.
What Actually Decides Your Citation Share
Citation share is downstream of citation distance: how many hops sit between your page and the place a number originates. An engine assembling an answer walks the reference trail behind a claim and tends to credit the page where the trail bottoms out. If your page restates a figure that lives somewhere else, you read as a hop on the way to the source, and the citation lands at the source, not at you. That mechanism is the subject of a separate piece on why ranking does not equal citation. The rule is enough here: credit lands where the trail bottoms out.
Being cited assumes being reachable first. When Ahrefs studied where cited URLs come from, 88% of the pages ChatGPT cites were taken directly from search, which means the same retrieval surface you already work still gates the door. Reachability gets you considered. Chain position decides whether you are the one credited once you are through it.
Which raises a prior question about any leaderboard: who was ever eligible to top it.
Why the Citation Leaderboard Is Mostly Pre-Decided
Picture one number's fate. You publish a figure you measured. Three blogs restate it without linking back to you, an engine resolves the claim to whichever of those pages reads cleanest, and the credit lands a hop short of the place the number was born. Multiply that by a corpus and you get the shape of the whole leaderboard.
Claim: Across 1,469 traced blog citations, only 15.0% reach a primary source, and the mean chain runs 1.08 hops, so most cited facts are borrowed restatements an engine can attribute to anyone. Source: LiquiChart SaaS Blog Citation Provenance Study Verified: 2026-06-14
The top of the leaderboard is locked before you enter it. When Ahrefs looked at the thousand pages ChatGPT cites most, 67% sat in categories ordinary outreach cannot touch: encyclopedias, homepages, educational domains, app stores. You are not going to outrank an encyclopedia for a definitional citation.
The contestable band sits underneath that ceiling, and inside it the origin slot is the one almost nobody fights for, because 15.0% of chains ever reach it. The crowded slots are the restating hops. The open slot is the source.
Reading those traces back, that empty seat is what surprised me: the origin almost nobody had claimed.
The Input You Control Is Owned vs Borrowed Numbers
The one input you control over your AI citation share is the ratio of owned to borrowed numbers across your pages, and unlike the share itself, you can count it today. Open your 20 most important pages and tag each load-bearing number by where it originates: yours, measured here, or borrowed from someone a hop upstream. That ratio is the leading indicator the percentage only reports after the fact. The dial is a count you can run this afternoon.
For a long time I ran that count by hand on our own posts: I would pick the number I least want an engine to credit to someone else, then chase it back to see whether it landed on us or on a page we had borrowed it from.
That is exactly what the Claims layer and Citation Provenance were built to read. Run a page through the Content Health Scanner and it extracts every statistical claim, then assigns each one a claim origin: a deterministic classification with five values.
Original when the number is backed by your own poll, chart, or confirmed research. First-party when you authored it but have not yet pinned it to data. Sourced when it points to a named outside source. Unattributed when it cites nothing. Unknown before classification has run.
No model guesses the label. The rules decide it. The same pass returns an Originality Score, the proportion of your claims that resolve to your own data rather than recycled assertions. You can scan a page for owned versus borrowed claims before any tracker could report a single answer.
Owning a number is also the move that earns the citation. The research behind generative engine optimization found that adding original statistics a reader cannot get elsewhere can raise visibility in generative answers by up to 40%. Turning a borrowed figure into one you own the number for is the work that follows the count, and it carries its own playbook.
How to Watch Citation Share Without Buying a Dashboard
Reading the number is cheap. Worth doing, on a slow cadence, and it needs no subscription: take a query your buyers actually ask, read the answer the engine returns, and note whether your domain is named among the sources. Do that across a fixed set of prompts and you have a defensible read on where you stand.
When you want to verify a single citation rather than a whole answer, the AI Citation Checker reads the live page behind a claim and tells you whether it states what you think it states, quoting the exact supporting sentence back. That confirms one source at a time. It does not score a share, and it was never built to. If the question is how your position moves across months rather than what it reads this week, watching position over time is the longitudinal cousin worth the attention.
Resist reading freshness as the lever it is not. Cited pages do skew newer: Ahrefs found AI assistants cite content averaging 1,064 days old against 1,432 for organic results, about 25.7% fresher. Newer on average, and still close to three years old at the mean. A recent date helps you get retrieved. What carries the citation is provenance.
Why We Do Not Sell a Citation Share Tracker
The dashboards will not say this part. LiquiChart does not track your citation share across competitors. There is no leaderboard in the product, and there is no version of one coming. I decided we would not build it. The percentage is a readout, and a readout is the wrong thing to wrap a dial around.
What living content infrastructure does instead sits at the input. Citation Provenance traces each statistic on your page through its reference chain and records where it bottoms out, whether at a primary source or at an aggregator standing in for one, so the Claims layer can tell you which of your numbers an engine can attribute to you and which point past you to whoever you borrowed them from. Living Content keeps the ones that are yours current as the underlying data shifts, so you stay the freshest path to the figure rather than a stale copy of it. Provenance position is the lever. Share is what the lever moves, on its own schedule, after you have set it.
Citation Share Follows Position
A scoreboard tells you the game already happened. The work that decided it happened earlier, on the pages where you chose to measure a number or borrow one. Share follows position, and position is set by how many of your load-bearing facts an engine can trace back to you instead of past you.
The slot that wins your share is open on almost every page you publish, sitting behind a borrowed number you could have measured yourself. Leave it borrowed, and it belongs to whoever you cited.