Skip to content
Help Centre
zicy.com
Try for free

Per-Prompt Results

What this is for, in one sentence: this is where you open a single tracked prompt and read the actual AI answers behind your numbers — the ground truth every dashboard metric is built from.

When to come here:

  • Whenever a number surprises you (“why did coverage drop?”)
  • When you want the exact wording AI uses about you
  • When you need evidence for a report: dashboards summarize, this screen shows receipts

We follow Sarah at MenuPilot, opening her prompt “Which software helps restaurants price menu items?”


  • Brand Mentioned. Whether, and how prominently, the answer named you. Good is a prominent mention; a single buried mention is labelled low prominence.
  • Position. Your place in the recommendation order in that answer. #1 or #2 is a headline recommendation; #4 or lower is an afterthought. Lower is better.
  • Mentions Frequency. How many times your name appears in that answer. More than one means the answer discussed you rather than name-dropped you.
  • Share of Voice (per answer). Your share of all brand mentions in that one answer.
  • Website Cited. Yes or No: did the answer link to your site. Being named and being linked are separate events.
  • Google Local Pack. Not eligible, Nothing found, Showing (with your position) or Not showing. It is Google’s map pack, not an AI answer.
  • Citation Sources. Sites the answer drew on, split into owned, earned and competitor.
  • YOU badge. Marks your own brand in the ranked Brand Mentions list.

Opening a prompt shows how each of the five AI platforms answered the same question — because they rarely agree.

one question, five different outcomes

For each engine you get five facts:

  1. Brand Mentioned — whether your brand was named, and how prominently. This isn’t just a blunt Yes/No: a single mention buried deep in an answer is labelled as the low-prominence mention it is, rather than turning a triumphant green “Yes” that reads like an endorsement.
  2. Position — where you appeared in the recommendation order. #2 means second brand named.
  3. Mentions Frequency — how many times your name came up in that answer.
  4. Share of Voice — your share of all brand mentions in that one answer.
  1. Website Cited (Yes/No) — did the answer link to your site.

Sarah’s spread is typical: strong on ChatGPT (#2, cited), barely present on Gemini (#4, one mention), absent from Perplexity and Google AI Overviews entirely. Per-engine differences like this are normal — each platform draws on different sources. They tell you where to focus, which is exactly the point.

Under the coverage figures, every prompt has a Google Local Pack row. It is a different signal from the five AI answers: it concerns Google’s own map pack, the three local businesses Google lists for a search.

The row reads Not eligible when the prompt isn’t a local search, Nothing found when the prompt is local but no map pack has been found for it (or it hasn’t been checked yet, or the check failed), and Showing with your position, such as #2, when your business is in the pack. If a pack was found without you in it, the badge reads Not showing.

Further down the page, a Google Local Pack panel explains the state in words. Local prompts are checked automatically when the prompt is analysed, and you can check again by hand. For a local prompt that has no result yet, it shows a Check Google Local Pack button; once the prompt has been checked, the button reads Re-check.

When a map pack is found, the panel lists the businesses Google shows, each with its position, website and address, and a You badge on your own business.

Zicy decides this from the wording of the prompt, not from the prompt’s Location setting. A prompt gets a map pack check in two cases:

  • It uses a local phrase, such as “near me”, “nearby”, “close to me”, “closest”, “open now”, “in my area”, “around me” or “walking distance”.
  • It says “in”, “near” or “around” a named place smaller than a country. For example, “What’s the best clinic for aesthetic treatments in Bangsar?” qualifies, and so would a city or state such as KL, Petaling Jaya or Selangor.

Wording that only names a country, such as “in Malaysia” or “in MY”, does not qualify. Write the place name with a capital letter: “in Bangsar” is picked up, but “in bangsar” is missed.

The Location setting is only used as the country to search in. A prompt set to a city but with no local wording still shows Not eligible. When the prompt names a place, Zicy searches Google from that place where it can, and from the country if it can’t.


The important habit on this screen is to read how strongly you show up, not just whether you show up. Being named once at the bottom of an answer a rival dominates is a very different result from being a headline recommendation — and Zicy now reflects that difference in the Brand Mentioned label itself, so a technical presence no longer masquerades as a win.

To judge a result fully, read three columns together:

  1. Brand Mentioned — are we in the conversation, and how prominently?
  2. Position — are we a headline recommendation (#1–2) or an afterthought (#4+)?
  3. Mentions Frequency — did the answer discuss us, or name-drop us once?

Sarah’s Gemini result (#4, 1 mention) is a presence, not an endorsement — and it’s labelled as the low-prominence mention it is. Her ChatGPT result (#2, 3 mentions, cited) is a real win. The screen makes the gap between the two legible instead of flattening both into the same green tick.


Open an engine’s row to read its complete answer — the exact text a real buyer saw. The page loads the numbers first and fetches each answer only when you ask for it, so a new row shows a Show AI answer button with the note that it loads the full answer this engine gave. Select it (or the row’s arrow, or Expand) and the answer appears, after a short Loading answer… wait. If it can’t be loaded, you see a message and a Try again button, and a row with no saved text says so. Once loaded, the answer stays available for the rest of your visit, and Collapse folds it back to a two-line preview. The answer text is shown clean, exactly as the buyer would read it — no raw citation tokens or markup cluttering the prose.

ChatGPT’s full answer, with the sources it cited

Three things to do here:

  1. Read how you’re framed, not just that you appear. “MenuPilot is a good fit for independent restaurants” and “MenuPilot also exists” both count as mentions; only one wins customers. This framing is what the Sentiment module measures at scale — here you see the raw material.
  2. Read who’s framed better. The brand described first, longest, and most concretely is winning this answer. What is AI saying about them that it isn’t saying about you?
  1. Check the Citation Sources list — the colour-coded list of every site this answer drew on: your own site (owned), third-party sites (earned), or rivals’ sites (competitor). If an engine consistently leans on one review site, that site is where your category’s answers are being decided. (The full source-type breakdown across all prompts lives in Citation Analysis, M-10.)

Copy Full Response / Download — use these to pull exact quotes into reports or share an answer with your team. What you copy or download is exactly the clean text shown on screen, so quotes paste in ready to use. Quoting AI verbatim is far more persuasive to a boss or client than any chart.


Below the answers sits the leaderboard for this prompt: every brand named across the engines’ answers, ranked by mention count.

who owns this question

This is the fastest way to see who dominates a specific question. Sarah’s list shows PlateIQ Labs at #1 with 7 mentions to her 6 — close, which makes this prompt a genuine battleground worth fighting for.

Finding yourself in the list. Your own brand carries a green YOU badge, so a mention count next to a look-alike name can’t be mistaken for yours — and look-alike brands display their domain so you can tell them apart. If a name you don’t recognise appears, open the full answer to see the context.


Google’s map pack is checked separately from the five AI engines. Zicy decides whether a prompt is a local search from its wording (see Which prompts get a map pack check) and checks local prompts automatically when the prompt is analysed. The result is stored per prompt.

You see it in three places. The Google Local Pack row under the coverage figures on this page; the Google Local Pack panel further down, with Check Google Local Pack or Re-check; and a Google Local Pack row inside a tracked prompt’s expanded breakdown on Prompt Manager, which also has a re-check button. The roll-up of every prompt with a map pack sits on the AI Visibility Dashboard.


  • Who can open it. Owners, members and viewers can open a prompt’s results.
  • Re-checking the map pack. A manual Local Pack check needs permission to use tools, which viewers do not have. Viewers can read the result but not re-run it.
  • Plan tier. No enterprise-tier plan is needed: the sidebar entry for Prompts, where you reach a prompt from, is not marked enterprise-only.
  • What spends allowance. Reading results spends nothing. The analyses behind them were counted on the Analysis quota when the prompt ran (see Prompt Manager). The Local Pack check code does not draw on that quota.

Why is my brand on some engines and not others? Each platform learns from different sources and updates at different times. Absence on one engine is a targeting insight, not a malfunction — see which sources that engine cites and earn your way into them.

An answer names a brand I’ve never heard of — why? AI answers reflect the market as AI sees it, which includes small players, adjacent tools, and occasionally brands from neighbouring categories. The Competitive AI Performance module (M-05) is where you decide who’s worth formally tracking.

Can I share one answer with a colleague? Yes — Copy Full Response for pasting, or Download for the file. The exported client report (M-15) also includes tracked-prompt results in summary form.

The answer text mentions us but Website Cited says No — contradiction? No: being named and being linked are different events. That gap — famous but never cited — is common, and closing it is what the Citation Analysis module and the Take Action and Content tools are for.


  • Zoom out to the scoreboard these answers roll up into: AI Visibility Dashboard (M-04).
  • See mentions by brand across ALL prompts: Competitive AI Performance (M-05).
  • Understand who gets cited and why: Citation Analysis (M-10).