Everything Cituna reports is downstream of one list: the buyer questions you track. Get that list right and the score, the competitor set and the content queue all describe your real market. Get it wrong and you will measure a market you do not sell into. Cituna proposes a starting list from your site, but the ten minutes you spend editing it is the highest-return work in the product. The same goes for competitors: the useful set is not the one your sales deck names, it is the one the engines actually recommend when a buyer asks. This guide covers writing good prompts, editing the tracked set, and reading the competitor screen without flattering yourself.
1. Write prompts the way a buyer types them
Track the question, not the keyword. "best ai visibility tool for a small agency" is a prompt; "ai visibility" is a category. Engines answer questions, and the answer either names you or does not.
Cover the buying journey rather than one stage of it: the problem question ("why is chatgpt not recommending my brand"), the category question ("best tools for tracking ai citations"), the comparison ("x vs y"), and the objection ("is x worth it").
2. Include the questions where you expect to lose
A tracked set full of prompts you already win is a flattering dashboard and a useless one. The prompts where a competitor is named instead of you are the ones that carry the work.
Add the two or three questions you dread hearing on a sales call. Those are usually the ones with the most commercial intent.
3. Edit the tracked set as your market changes
Open the Prompts tab to add, edit or remove tracked prompts at any time. A prompt you remove stops being checked and stops counting toward your score.
All six engines are checked daily on every plan, so a prompt added today starts producing a trend within the week.
4. Read who is cited instead of you
The Competitors tab lists the brands engines name for your prompts, and how often. This set is discovered from real answers, not assumed from a market map, so expect surprises: often a publisher, a marketplace or a review site rather than a direct competitor.
When a review site or a listicle keeps appearing, that is not a competitor to beat. It is a page to be included on, which is a different and usually cheaper job.
5. Follow the citations to the pages doing the work
Open Citation sources to see the specific pages engines cite when they answer your prompts. This tells you the format currently being rewarded: a comparison table, a pricing page, a forum thread, a documentation page.
Match the format before you try to beat the content. A better essay rarely displaces a table.
Troubleshooting
How many prompts should I track?
Fewer, better ones. Ten questions you would recognise from a real sales conversation produce a more actionable picture than fifty generated variations, because every prompt you add dilutes the score with a question nobody asks.
A competitor appears that we never compete with.
That is a finding, not noise. It means the engines consider them an answer to your buyer question, which is what a buyer using ChatGPT will see too. Check which of their pages is being cited before deciding whether to act.
Our own brand name is in a prompt. Does that inflate the score?
It can. Branded prompts are worth tracking to catch misinformation about you, but keep them separate in your reading from unbranded prompts, which are the ones that measure discovery by people who do not know you yet.
Why do results differ between engines for the same question?
They use different retrieval, different indexes and different training data, and some search the live web while others answer from memory. Cituna labels how each answer was produced, so an engine answering without web access is never counted as evidence that your page was ignored.