Reading the Overview dashboard

Overview is the first screen of your project: three headline numbers, a trend chart, and a competitor comparison. Everything on it is computed from the AI answers Maya collected for your tracked prompts.

The three headline KPIs

• Mention Rate — the share of AI responses that mention your brand: responses mentioning you ÷ total responses × 100. Every provider response (ChatGPT, Gemini, Claude, Perplexity, …) counts separately. The card shows a "Low data" note until enough responses are collected, and the trend arrow needs at least ~20 responses at both ends of the comparison window.
• Sentiment — how answers talk about your brand when you appear, scored 0–100. Only brand-mentioning answers feed this number.
• Share of Voice — your slice of the conversation: each response adds +1 to your tally if you're mentioned and +1 for every unique radar competitor named in it. SoV is relative, so it can fall even when your mentions grow — it means competitors grew faster.

Filters

• Country — single select; AI assistants answer differently per market.
• Provider — multi-select; pick providers and hit Apply. Useful to isolate a provider-specific drop.
• Labels / Non-branded — the Non-branded toggle restricts everything to category questions (where your name is not in the prompt). This is the honest view of discovery visibility: branded prompts almost always mention you and inflate the averages.

Filters combine (AND) and apply to the whole page, so the KPIs, chart, and competitor table always describe the same slice.

The main chart

Switch the metric (Mention Rate / Sentiment / Share of Voice) and the mode:

• Competitor — your line plotted against your radar competitors.
• Solo — just your trend.

A summary headline above the chart states the net movement in the selected window.

Competitor Comparison

The table ranks you and your radar competitors by Mention Rate for the current filter slice. Competitors are managed on the Competitors page — adding or hiding a competitor there changes this table and the Share of Voice math.

Citation Share

A stacked chart of who owns the cited sources in your answers: your brand, each radar competitor, and "Other". It defaults to non-branded questions (the competitive view) and can be switched to all questions, and between domain-level and URL-level.

When the dashboard is empty

A new project shows "Analyzing Your Prompts…" while the first runs complete — usually minutes, up to the first daily cycle. If it stays empty, check that prompts exist and are active in Prompt Lab.

How to read it without fooling yourself

1. Judge discovery visibility with the Non-branded filter on.
2. A one-day dip is usually noise; compare week over week.
3. If a drop is isolated to one provider, investigate that provider's answers before concluding anything about your brand.