Features

What Citerva measures, and what it deliberately does not claim

Mention rate, per provider
The share of answers that named you, out of the answers that actually came back. Reported overall, per assistant and per topic — never averaged into a single score, because a brand at 30% overall but 0% on one assistant has a different problem from one evenly at 30%.
Competitor share of voice
Every competitor scored over exactly the same answers as you. An answer naming you and two rivals gives you one point of three, so appearing alongside everyone is not the same as winning.
Citation and source analysis
Which domains assistants pointed at, whether any were yours, and whether the provider declared a source or the link merely appeared in prose. The two are counted separately, because a declared source is a much stronger signal.
Prompt-level evidence
Every answer stored verbatim, against the prompt that produced it. Each mention and citation traces back to the sentence it was read from, so no figure has to be taken on trust.
Absence, separated from unknown
A prompt where an assistant answered and did not name you is absence, and actionable. A prompt where nothing answered is unknown, and is not counted as either. Most tools conflate the two.
Evidence-linked findings
Each finding names the counts behind it and the prompts it came from. Nothing claims that a change will make a model recommend you — findings point at what was observed and what is worth investigating.
Scheduled scans and trends
Weekly or daily re-scans so you read a direction rather than a sample. Comparisons are only drawn between scans that asked the same prompts of the same providers; changing either marks a break in the line instead of a misleading slope.
Team workspaces
Multiple projects per workspace for agencies, role-based access, and tenant isolation enforced in the database rather than in application code.

What Citerva does not do

Worth knowing before you buy, not after.

  • It does not make AI assistants recommend you

    There is no lever here that changes a model's output. This is measurement. Anyone selling guaranteed AI recommendations is selling something they cannot deliver.

  • It does not report a ranking position

    AI assistants do not publish a ranked index, so there is no position three to hold. What exists is the share of answers you appear in, and the order you tend to appear in when you do.

  • It does not turn one scan into a verdict

    Ask the same question twice and a model may name different brands. A single scan is one sample. We show the counts, flag small samples, and will not label a one-run difference as movement.

  • It does not attribute cause

    If your mention rate rises after a content change, that is correlation. Model updates and index refreshes move these numbers on their own, and we will not claim your change was the reason.