Understanding the metrics in the prompts table

Under Prompts → Alle Prompts (All prompts) in the Sichtbarkeit (Visibility) sidebar group, KI Score shows for each of your prompts how often AI answers name your brand, cite your website as a source, and bring up your tracked competitors. This article explains every column, the filters, and a prompt’s detail panel.

All rates refer to the answers that could be evaluated in the selected period. Failed answers are not counted. You can also see a column’s explanation right in the dashboard by hovering over its header. Hovering over a cell shows the number of answers behind it.

Mentions: rate, competitors, and gap

• Erwähnung % (Mention %): the share of completed answers with text that name your brand name or an alias.
• Am häufigsten erwähnter Wettbewerber % (Most-mentioned competitor %): the name and mention rate of the single tracked competitor mentioned most often for this prompt.
• Wettbewerber erwähnt % (Competitor mentioned %): the share of answers that mention at least one tracked competitor. You can show this column via Spalten (Columns) when needed.
• Erwähnungslücke (Mention gap): your mention rate minus the most-mentioned competitor’s mention rate, in percentage points. A negative value means that, for this prompt, that competitor is mentioned more often than your brand.

In the dashboard glossary, you’ll find the gap under Erwähnungslücke, and also by searching for “Mention Gap”.

Citations: rate, competitors, and gap

• Zitation % (Citation %): the share of answers with evaluable source information that cite your own domain, including subdomains, as a source.
• Meistzitierter Wettbewerber % (Most-cited competitor %): the name and citation rate of the tracked competitor whose domain is cited most often.
• Wettbewerber zitiert % (Competitor cited %): the share of answers that cite at least one competitor domain. This column is hidden by default too.
• Zitationslücke (Citation gap): your citation rate minus the most-cited competitor’s citation rate.

The competitor citation columns require that you have added a domain for your competitors. Competitors without a domain are left out there. A mention only means the AI talks about your brand. A citation means it lists your website as a source.

Position and brand sentiment

• Ø Position (Avg. position): your brand’s average position among all providers named, where 1 means named first. Only answers that name your brand with a known position count.
• Tonalität (Brand sentiment): how the AI classifies statements about your brand, positive (praise or a recommendation), neutral (a factual mention), or negative (criticism or advice against it), shown as a bar. Your brand receives one classification per answer. The bar appears from five classified answers. The article What brand sentiment tells you explains how to read it.

The separate recommendation rate column was removed from the table on October 26, 2026. How often the AI recommends or praises your brand shows up as the positive share of the brand sentiment.

Measurement country, tags, wordings, and answers

• Messland (Country): the countries the prompt is measured from, shown as flags. A prompt can have up to 10 measurement countries. Each country is measured separately and costs credits like a prompt of its own.
• Tags: in front of the tags you see B (Branded) if the question names your brand, otherwise N (Non-branded). KI Score derives this from the prompt text. Clicking the cell opens Typ (Type) and Tags, where you can correct the type by hand if the detection doesn’t fit.
• Formulierungen (Wordings): the number of wordings under which the prompt was measured. As long as you don’t change the text, this shows 1. After an edit, the metrics also include answers to earlier wordings. The column is hidden by default.
• Antworten (Answers): the completed answers in the selected period.
• KI-Nachfrage (AI demand): the estimated monthly AI demand for the topic KI Score automatically matches to the prompt, separately for each measurement country. It isn’t a count of queries for this exact prompt and is never summed across prompts.

Period, filters, and the total row

Above the table, you choose which data the metrics cover:

• Zeitraum (Date range): Letzte 7 Tage (Last 7 days), Letzte 14 Tage (Last 14 days), Letzte 30 Tage (Last 30 days, the default), Gesamter Zeitraum (All time), or any range in the calendar. With vs. Vergleich (vs. baseline) you can also pick a comparison period, such as Vorheriger Zeitraum (Previous period) or Gleiche Tage im Vormonat (Same days in the previous month). The rates, the gaps, and the average position then show their change; brand sentiment doesn’t.
• Prompt-Status (Prompt status): a fixed filter next to the date range with Alle (All), Aktiv (Active, the default), and Pausiert (Paused).
• Filter hinzufügen (Add filter): among others Prompt-Suche (Prompt search, for prompts whose text contains, or doesn’t contain, a term), Modell (Model), Thema (Topic), Tags, Messland (Country), Nennt Marke (Mentions brand: Ja, Yes, or Nein, No), Wettbewerber erwähnt (Competitor mentioned), Wettbewerber zitiert (Competitor cited), and thresholds for Erwähnungsrate (Mention rate), Zitationsrate (Citation rate), and Ø Position (Avg. position). With Modell, KI Score calculates all metrics only from the answers of the selected models.

The table is grouped by topic. The Gesamt (Total) row summarizes the prompts shown. For the rates, it counts all answers of the prompts shown together, so a prompt with more answers counts more. For brand sentiment, it adds up the classified answers. For the most-mentioned and the most-cited competitor, each prompt contributes its own strongest one, which can differ from prompt to prompt. Answers and wordings are added up; AI demand isn’t. The footer shows how many prompts are displayed.

Note: KI Score always calculates the metrics with your current brand, alias, competitor, and domain details. If you change them, the values for answers already stored change too.

A prompt’s detail panel

1. Click a prompt’s row in the prompts table.
2. Read the cards and charts in the panel that opens.
3. Click a card’s info icon if you want to read how it’s calculated in detail.

The panel evaluates this prompt’s measurement runs in the period you selected above the table. The period is shown above the cards. The Übersicht (Overview) tab shows the following cards; the Fanouts tab shows the web searches the AI models ran for this prompt only.

• Auswertbare Antworten (Evaluable answers): the number of completed answers with text or sources that could be evaluated. From here, you can go to the matching answers in Responses.
• Getrackte Quellen (Tracked sources): the number of distinct domains the AI names as sources in these answers, your own and others combined.
• Zugeordnete Zitationen (Attributed citations): the share of answers with known source attribution that cite your domain or one of its subdomains.
• Erwähnungsraten im Vergleich (Brand mention rates compared): your mention rate next to the mention rates of up to five most-mentioned tracked competitors. Since an answer can mention several brands, the rates don’t need to add up to 100 percent.
• Sichtbarkeit je Anbieter (Visibility by provider): the mention rate for each AI provider.
• Ø Nennungsposition je KI-Anbieter (Avg. mention position by AI provider): how early your brand appears among the relevant brands for each AI provider, with the number of answers behind it.
• Erwähnungsrate im Zeitverlauf (Mention rate over time) and Zitationsrate im Zeitverlauf (Citation rate over time): the trend across the measurement runs in the period, one point per run.
• Formulierungen (Wordings): the current prompt text and earlier versions, each with its number of answers.