Read your visibility dashboard
The dashboard answers one question: are AI answer engines recommending your brand, and is that getting better or worse? Read it top to bottom: Share of Voice, trends, leaderboard, then the AI engine breakdown, and finish by drilling into a single prompt to see the actual answers. The numbers are probabilistic, so read trends, not single runs.
Before you start
You need a locked prompt-set version that has run once. If you just created your project, run the prompts first, see create a project. A trend only appears once you have two or more dated runs, so turn on a schedule early.A tour of the dashboard
Start with Share of Voice
The headline metric is Share of Voice: your average share within non-branded answers that name at least one direct brand. It answers the category question: when buyers do not ask for you by name, how much of the answer belongs to you compared with the brands that actually appear?
Read the trend line, not the point
The headline uses completed weekly runs, giving each week equal weight in the selected period. Below it sits the weekly trend for the same prompt-set version. A single run can swing because AI answers vary, so the shape over the version's weekly runs is what matters. A sustained drop from 31% to 18% is a real signal worth investigating.
Check the leaderboard
The leaderboard ranks every direct brand found in the selected answers. Your rank and the total number of observed brands stay visible above the scroll, even when your row is farther down. The citation leaderboard does the same for your site among cited domains.
Open the AI engine breakdown
AppearIn AI tracks ChatGPT, Gemini, Perplexity, Google AI Mode, and Google AI Overviews separately, and they disagree often. The breakdown shows your mention rate, citation rate, and Share of Voice on each. A healthy rollup can hide a brand that's invisible on Perplexity.
Drill into a single prompt
Click any prompt to read the saved answers from each AI engine for that run. This is the ground truth behind every metric: you see exactly how you were described, who was named ahead of you, which pages were cited, and whether any hallucination flags fired.
A new set starts a new baseline
When the next monthly version locks, its metrics start from a fresh set-scoped baseline. Earlier results remain attributed to their original version. Use the Period control to revisit them. The Compare option shows direct deltas only when both versions use the same prompts and run settings; otherwise, inspect each period separately.What to look at first: coverage gaps
Before optimising anything, find your coverage gaps, prompts where you are barely mentioned or absent entirely. These are the cheapest wins, because going from absent to present on a non-branded prompt directly improves organic visibility. Filter for losing prompts, look at the bottom of the list, and read those answers to learn what the AI engines recommend instead.
Which metric should I watch?
The dashboard surfaces several metrics and they answer different questions. Pick the one that matches what you're trying to learn.
| Metric | Answers the question | Best for | Watch out for |
|---|---|---|---|
| Mention Rate | How often am I named at all? | Finding coverage gaps and tracking raw presence. | Ignores ranking, you can be mentioned last and still count. |
| Share of Voice | How do I stack up against competitors? | Competitive benchmarking by AI engine. | Uses only direct brands found in eligible answers. |
| Citation Rate | Whose pages do AI engines link as sources? | Spotting which sites earn an AI engine's trust. | Your own pages cited less than third-party ones is normal. |
| Sentiment | When I'm named, am I described well? | Catching negative framing even when visibility is high. | Needs enough mentions to be meaningful; sparse data is noisy. |
For the full definitions behind these, mentions, citations and sources and sentiment each have their own concept page.
Worked example
Acme tracks the prompt "best project management tools for startups". The blended dashboard reads:
- Share of Voice 31%, down from 42% two weeks ago, the trend, not the single weekly reading, is the alarm.
- Leaderboard: Beta leads Share of Voice at 48%, Acme second at 31%.
- By AI engine: Acme is solid on ChatGPT (Share of Voice 52%) and Gemini (44%), but only 9% on Perplexity.
Drilling into the prompt on Perplexity, Acme reads the saved answer: it lists Beta and two review-site favourites, and never names Acme. The blended rollup looked fine; the AI engine view exposed that Perplexity effectively doesn't know Acme exists. That is the coverage gap to fix first.
What good looks like
- You read the trend over weeks before reacting to any single run.
- Your Share of Voice is rising across several completed weekly runs.
- No single AI engine is a dead zone, coverage is spread, not lopsided.
- When you spot a moving number, you open the prompt and read the actual answer instead of guessing why.
A simple reading cadence
Weekly, glance at the headline trend and the leaderboard. Monthly, review each AI engine and re-read your weakest prompts and prepare the next set if your priorities changed. When a new competitor appears, open the underlying answers before treating it as a durable shift.
Common mistakes
Reading a single run as truth. AI engine answers vary between runs even with the same prompt. One bad run is noise; a sustained slide is signal. Always check the trend before you act.
Letting the blended view hide AI engine reality. Averaging across AI engines smooths over the brand that's invisible on Perplexity. A healthy rollup can mask a dead zone, always open the AI engine breakdown.
Next steps
The metrics behind presence, citations, reputation, and competitive standing.
Why ChatGPT, Gemini, Perplexity, Google AI Mode, and Google AI Overviews disagree, and why we track them apart.
A diagnosis path for when the trend turns down.