Citation share
The percentage of all sources cited across a set of AI answers that belong to your domain. The AI-era equivalent of ranking share.
In depth
What it really means
Citation share counts links, not name-drops. When Perplexity answers a question and lists eight sources, citation share asks how many of those eight are yours. That makes it the most concrete AI visibility metric available, because a citation is a discrete, countable, usually clickable thing.
It behaves differently from search ranking in one important way. There is no position one. An answer citing six sources gives all six some visibility, so the game is getting into the set rather than beating everyone to the top. That makes the metric less winner-take-all and more like share of voice.
Citation share = (your citations across the query set / total citations across the query set) x 100
Fix the query set before you measure anything. A shifting query set makes the trend meaningless, which is the most common way this metric gets misused. Run each query at least three times and average. Assistants are non-deterministic and single readings swing wildly. Benchmark against your five real competitors, not the whole web. Wikipedia and G2 will dominate the raw denominator and tell you nothing actionable.
Pros & cons
Pros
- Citations are countable and often clickable, so the metric maps to something real.
- It attributes cleanly to specific pages, which tells you what to write more of.
- It moves faster than rankings, so content changes show up in weeks.
Cons
- Measuring it properly means running hundreds of prompts, which is slow by hand and expensive by tool.
- Non-determinism adds noise that hides small real changes.
- A citation does not guarantee a visit, so the metric can rise while sessions fall.
Common mistakes
- Changing the query set between measurement periods, which destroys comparability.
- Measuring once and treating it as a baseline. One run is noise.
- Counting Wikipedia and review aggregators in the competitive set, which drowns out the brands you actually compete with.
- Comparing citation share across assistants as if it were one number. They cite very differently and should be tracked separately.
- Optimizing for total citations rather than citations on queries with buying intent.
Best practices
- Define 30 to 50 queries that mirror how your buyers actually ask, and freeze the list.
- Run each query three or more times per assistant and average the result.
- Track per assistant and per query cluster, never as one blended figure.
- Attribute every citation to a URL so you know which pages earn them.
- Weight the query set toward commercial intent, where a citation is worth something.
FAQs
What is citation share?
The share of all sources cited across a defined set of AI answers that point to your domain, measured against competitors. It is the closest AI-era equivalent to ranking share.
How do I measure citation share?
Fix a query set, run each query several times across each assistant, log every cited URL, then divide your citations by total citations. Track it per assistant, because they cite very differently.
How is citation share different from share of model?
Citation share counts linked sources. Share of model counts brand mentions, linked or not. You can be mentioned constantly and cited never, which usually means the models know you but do not trust your site as a source.
What is a good citation share?
There is no absolute benchmark. It only means something against your five real competitors on your query set, and the trend matters far more than the level.
Why does my citation share swing so much?
AI assistants are non-deterministic, so the same query returns different sources on different runs. Average at least three runs per query, or you are measuring noise.
Keep reading
Related on LymLyt
Beyond LymLyt
Further reading
Want this working on your site?
We build the content behind the term, ranked in search and cited by AI.
Book a 30-min call →