Documentation
Back
  • Plans, limits and what a scan costs
  • What Citerva measures
    • Setting up your first project
  • Running scans and reading the evidence
  • How every number is calculated

Setting up your first project

How to describe your product, choose competitors and topics, and get a prompt set worth scanning.

A project is one brand you are tracking. An agency usually runs one project per client; a single company usually runs one per product.

Describe the product properly

The description is what prompt generation works from, so a vague description produces vague prompts. Write it the way you would explain the product to a buyer, not the way you would write a meta description.

Good:

An applicant tracking system for early-stage startups: fast to set up, opinionated defaults, structured interview scorecards without an enterprise rollout.

Less useful:

The leading AI-powered recruitment platform for modern teams.

The second one could describe forty products, so the prompts generated from it will be generic too.

Add your domains

The primary domain decides whether a cited link counts as yours. Subdomains are matched automatically, so docs.acme.com counts as owned if your domain is acme.com.

Add any documentation, help centre or blog that lives on a different apex as an extra domain — otherwise a model citing your own help centre will be counted as a third-party citation.

Add alternate names

If a model writes your legal name, an old name, or a common abbreviation, that is still a mention — but only if Citerva knows to look for it. Add each one on its own line.

Matching already handles case, accents and punctuation, so Acme.io, Acme-io and acme io are all the same name and do not need separate entries.

Very short names are matched more cautiously. A three-letter brand that is also an ordinary English word will be found, but scored at lower confidence, because there is no reliable way to tell the two apart from text alone.

Choose competitors deliberately

Competitors are scored over exactly the same answers you are, which is what makes the comparison meaningful. Without them, Citerva can tell you whether you appear but not who is taking the space you are not.

Pick the brands you actually lose deals to. Adding a competitor changes the share of voice denominator, so a stable list makes trends comparable over time.

Choose topics a buyer would recognise

Topics drive the questions. Phrase them as problems, not as keywords.

  • Good: hiring for early-stage startups, interview scheduling for remote teams
  • Less useful: best ATS software 2026

The second is a search query, not something a person asks an assistant.

Review the prompt set

Citerva generates the questions and shows them to you before anything is sent. Generating is free; asking an AI provider costs money, so this approval step is real rather than a formality.

Two things to look for:

Do these read like a buyer? Edit anything that sounds like SEO copy. You are measuring the questions people actually ask.

Is anything naming your brand? A question that names you will almost always produce a mention of you, so it measures nothing about discovery. Citerva marks these and excludes them from mention rate and share of voice automatically — but it is worth knowing which ones they are, because they answer a different and also useful question: what does a model say about us once it already knows who we are?