How to Read a Single Prompt's Performance in AmICited

Click any prompt in the Tracked prompts table and you land on its detail page: a full breakdown of how your brand performs for that single question. Before you read the numbers, it helps to understand what they actually represent.

Quick Steps

• Click a prompt row in the Prompts table to open its detail page.
• Check the four overview cards: Visibility, Avg position, Citations, and Engines cited.
• Start with Engines cited: a low ratio like 3/4 flags a specific engine gap to investigate.
• Read Visibility next to Avg position to see whether you’re absent or just under-ranked.
• Use Process now to re-run the prompt on demand after publishing new content targeting it.

What is prompt-level AI visibility?

A “prompt” in AmICited is a specific question a real customer might type into ChatGPT, Perplexity, Gemini, or ask via Google AI Overviews: things like “best project management software for small teams” or “how does [category] pricing work.” Unlike traditional keyword rank tracking, where one URL either ranks or doesn’t for a search term, an AI-generated answer is a synthesized response that can cite several sources, quote competitors by name, or mention no one at all. That’s what makes AI search visibility fundamentally different to monitor: you’re not tracking a position in a list of ten blue links, you’re tracking whether, and how prominently, your brand shows up inside a generated paragraph.

This is why prompt tracking exists as its own discipline. Each tracked prompt is run repeatedly against one or more AI engines on a schedule, and every run is logged as a data point: did the engine cite you, where in the answer did the citation appear, what sources did it pull from, and what did competitors get instead. A single prompt’s performance page is the rollup of all those runs into one scorecard, so you can judge your standing for that exact question rather than guessing from an aggregate score across your whole prompt set.

Two ideas do most of the work here. The first is citation position: where in the AI’s answer your brand shows up relative to everyone else mentioned. A citation that appears first, or is named in the opening sentence, carries more weight with the reader than one buried in a list of “other options” near the end, even though both technically count as a citation. Citation position is the AI-search equivalent of ranking #1 versus ranking #9 in classic search. The second is AI share of voice: of all the citations a prompt generates across its runs, what percentage are yours versus your competitors’. A prompt can have healthy citation frequency for you while your AI share of voice is still low, because two or three competitors are being cited more often or more consistently. Reading a prompt’s detail page well means holding both ideas, presence and prominence, in view at the same time, because either one alone can be misleading.

Why does any of this matter beyond curiosity? Because prompts are the new keywords. The prompts your buyers actually type are the units of demand in AI search, and each one either sends that buyer your name or a competitor’s. A prompt detail page turns an abstract worry (“are we visible in AI?”) into a concrete, per-question answer you can act on: fix content, request the citation-worthy source, or accept that a given prompt simply isn’t winnable yet.

Where to find it

From the Prompts table, click a prompt row. The detail page opens with a ← All prompts link, the prompt text as the title, its status badges (e.g. CITED, DAILY, country, tag), and the action buttons Generate article, Compare a URL and Export on the right. The status badges are worth reading before the numbers: CITED tells you at a glance whether your brand has ever appeared for this prompt, DAILY (or whatever cadence is set) tells you how fresh the underlying runs are, and the country tag tells you which market’s AI results you’re looking at: citation behavior can differ noticeably between, say, US and UK runs of the same question because the engines draw on different regional sources.

The overview stats

The four core cards summarize your standing for this prompt:

• Visibility: the share of this prompt’s citations that are yours (e.g. 7%). This is the prompt-level version of your overall AI visibility score, scoped to just this one question.
• Avg position: your average citation position across the answers where you appear (lower is better). This is calculated only from runs where you were actually cited, so a strong average position with low overall visibility usually means “when we do get cited, we rank well, we just don’t get cited often enough.”
• Citations: how many citations you earned, out of how many runs. This is your raw hit rate before position is factored in at all.
• Engines cited: how many of the tracked engines cited you (e.g. 3/4), and how many ran the prompt. Because ChatGPT, Perplexity, Gemini and Google AI Overviews pull from different indexes and weight sources differently, it’s entirely normal to be strong on one engine and invisible on another for the exact same question.

For prompts with search data, you’ll also see Search volume, CPC and Competition cards: the same market context shown in the prompts table, so you can judge how valuable winning this prompt would be. Search volume tells you how many people are asking some version of this question every month (via the underlying keyword data), CPC signals how commercially valuable advertisers consider the topic, and Competition tells you how contested the space already is. A prompt with high search volume and low citations-to-date is usually your best opportunity: real demand exists, and you’re not yet capturing it.

How to use it

1. Start with Engines cited. 3/4 means one engine isn’t citing you at all: a clear, specific gap. Check which engine is missing; if it’s consistently the same one across several prompts, that points to a systemic issue (e.g. content not indexed where that engine sources from) rather than a one-off.
2. Read Visibility next to Avg position. Appearing often and near the top is the strong position; frequent but low-ranked means you’re being out-cited: your content is being pulled in, but a competitor’s source is winning the more prominent citation position within the same answers.
3. Use Process now (next to the header stats) to re-run the prompt immediately instead of waiting for the schedule. This is useful right after you’ve published or updated content targeting this prompt: you don’t have to wait for the next scheduled run to see whether the change moved the needle.
4. Cross-check against search volume. A prompt with strong search volume but weak visibility and citations is a prioritization signal: it’s worth more editorial effort than a low-volume prompt with the same weak numbers, because fixing it reaches more of the audience actually asking that question.
5. Watch the trend, not just the snapshot. A single run can be noisy (AI answers vary run to run even for an unchanged prompt), so treat one weak result as a data point, not a verdict, until you’ve accumulated enough runs to see a pattern.

Everything below the stats (brand mentions, the actual AI responses, cited sources, fan-out queries) explains why these numbers look the way they do. The brand-mentions section shows the exact sentences where you were named (or where a competitor was named instead), the response history lets you read the full generated answer text engine by engine, cited sources shows which pages the AI actually pulled from to build that answer, and fan-out queries reveal the related sub-questions the engine generated internally while researching, often a preview of adjacent prompts worth tracking brand mentions in ChatGPT and other engines for.

Once you’ve read a handful of prompt detail pages this way, patterns start to emerge across your whole tracked set: certain engines favor certain source types, certain prompt phrasings trigger a citation and others don’t, and certain competitors show up consistently in the citation position you want. That’s the point at which a single prompt’s scorecard turns into a strategy: you can prioritize which prompts to target for new content, which existing pages need restructuring to be more citable, and which engines deserve extra attention because your AI rank tracker shows you’re losing ground there specifically. If you’re still building out which questions to track in the first place, it’s worth pairing this page with a broader look at how to measure AI search visibility across your full prompt library, so individual prompt wins add up to a defensible position in AI search overall rather than a handful of disconnected good results.