Visibility Tracking
See how often each AI engine mentions your brand — measured with multi-run sampling and broken down by engine, topic, and prompt.
Video walkthrough — coming soon
What the Visibility tab shows
The tab opens with three hero cards, followed by your trend over time, the engine table, topic visibility, the full response table, and gaps. A filter bar at the top applies to almost everything: Last 7 / 30 / 90 days or All time, an Unbranded toggle, and a CSV Export.
- AI Health Score — the overall score and grade, with a plain-English read on where you stand (from "Top 10%" down to "Bottom 20%").
- Mentioned Responses — "Mentioned in X of Y AI responses", split into branded and unbranded prompts.
- AI Reach — how many of the tracked engines mention you at all, with a per-engine ring.
The engine table
One row per engine — ChatGPT, Claude, Perplexity, Gemini — with columns for Mention Rate, Mentions, Traffic, and Trend.
- Mention rate — the share of that engine's scanned responses that mentioned your brand. Colour-coded: green from 60%, amber from 30%, red below.
- Mentions — the response count behind the rate.
- Traffic — currently shows "—": per-engine traffic isn't tracked yet, and Upranke shows a dash rather than an estimate.
- Trend — whether the engine is up or down against your recent average.
Below the main engines, a Google AI Surfaces section adds AI Overview and AI Mode rows, with their own refresh button and 15-day cooldown — covered in AI Overview & AI Mode.
Multi-run sampling
AI answers aren't deterministic — the same question can produce different answers minutes apart. So each prompt runs three times per engine, and a prompt's rate reflects how many of those runs mentioned you. A single lucky answer can't fake a trend, and every result carries a confidence level based on how consistent the runs were.
A measured 0% is not "no data". 0% means the engine was scanned and didn't mention you — a real result you can work on. Surfaces that have never been scanned show "Not tracked yet" instead. Upranke keeps the two strictly separate.
Topic and competitor heatmaps
Two grids break the same data down further: engine × topic (which engines mention you on which topics) and topic × competitor (who owns each topic's answers). This is where a flat "34% overall" turns into "strong on integrations, invisible on pricing".
Use the Unbranded toggle. Prompts that don't name your brand are the hardest — and most valuable — to win. Filtering to unbranded shows whether AI recommends you to buyers who've never heard of you.