The metric
AI Coverage
Some AI answers list the sources they used. Others answer without naming anyone. AI Coverage is the share that do, engine by engine, and it says whether there is a door at all.
Definition
AI Coverage is the share of an AI engine's answers that name at least one source. An engine at 100% shows its sources every time it answers; an engine at 50% does it in one answer out of two, and the rest is written without crediting anyone. It is measured per engine, over a stated set of questions and a stated period.
How the number is built
A share of answers, and nothing else: no constant to calibrate, no weighting to arbitrate. What it counts and what it leaves out is the whole of it.
- Every answer is in the denominator
- Answers that name nobody count too. They are the point of the metric: leaving them out would measure only the answers that already cite, and the number would always read close to 100%.
- Only the sources the engine credits
- A page the engine cites in its answer counts. A page it read without crediting does not, because engines differ in whether they expose those at all, and counting them would make one of them incomparable to the rest.
- One number per engine
- Engines differ far more from each other than a single engine does from one week to the next. A pooled figure would average away the only distinction that changes what you do.
What it tells you
AI Coverage is a measure of opportunity. It says how often an engine opens its answer to the outside at all, which sets a ceiling on everything else you might do there.
Read it as a ceiling
On an engine that cites in half its answers, half the questions your market asks are settled without any source being named. The other half is where sourcing work can land.
Compare engines, not periods
Its level depends on the questions asked, so it is read over a stated set of questions. Within one survey every engine sees the same questions on the same day, so a difference between engines is a difference between engines.
Watch it move
An engine can open up or close down from one month to the next. Epovest records the figure on every survey, so it is a curve rather than a single reading, and a change is attributable to the engine that produced it.
AI Coverage says whether. AI Density says how many.
On its own, AI Coverage tells you an engine cites, not how much room there is once it does. An engine may cite in every answer and name only two sources. The two numbers multiply into the count of distinct sources an answer carries, which is the figure to compare across engines.
Common questions
What is AI Coverage?
AI Coverage is the share of an AI engine's answers that name at least one source. An engine at 100% shows its sources every time it answers; an engine at 50% does it in one answer out of two, and the rest is written without crediting anyone. It is measured per engine, over a stated set of questions and a stated period.
How is AI Coverage computed?
It is the number of answers naming at least one source, divided by every answer the engine produced on those questions in that period. Answers that name nobody stay in the denominator, since they are exactly what the metric is there to count. Only pages the engine credits in its answer are counted.
How does AI Coverage differ from AI Density?
AI Coverage says how often an answer cites anyone; AI Density says how many distinct sources it names when it does. They are independent by construction, because AI Density counts only over answers that cite. Multiplied together they give the average number of distinct sources an answer carries.
How does AI Coverage differ from AI Authority?
They describe different objects. AI Authority is a property of a SOURCE: it says how much a given engine leans on a given site, on a scale where 100 is the site that engine cites most. AI Coverage is a property of the ANSWER: it says whether the answer cites anyone at all. One tells you which source to work on, the other how often there is a source to be.
Does AI Coverage depend on the questions being asked?
Its level does, so it is always read over a stated set of questions and a stated period, and Epovest reports it per engine on the questions you actually track. What it isolates is the engine: every engine in a survey sees the same questions on the same day, so a difference between engines is a difference between engines.
Make AIs recommend you
Create a tracker in minutes: your prompts, your engines, your cadence. First checks within 24 to 48 hours.
No subscription. Credits from $10, valid 12 months. Pause anytime: the history stays yours.