What AI Visibility Metrics Can Semji Track?

Semji tracks four core AI visibility metrics — AI Visibility Score, mention rate, citation rate, and average mention position — and breaks each of them down by AI platform (ChatGPT, Google AI Overviews, Perplexity, Gemini, Claude, DeepSeek) and by topic. The metrics live in an "AI Visibility" view that reports on a rolling window, shown in Semji's own product example as "Last 90 days · weekly." If you need to judge whether your brand is actually showing up in AI answers — and whether that's improving — these are the numbers to watch.

The four metrics, and what each one tells you

Mention rate vs. citation rate — the distinction that matters

These two are easy to conflate, but they answer different questions:

• Mention rate = "Is the AI talking about us?" A brand can be named in an answer without being used as a source.
• Citation rate = "Is the AI pointing to us as the source?" This is the stronger signal, because it means your page was used to construct the answer.

Semji's own example frames the gap directly: a citation rate that trails the mention rate means you're "linked as a source, not just named" — i.e., you're being referenced less often than you're being talked about. If citation rate is your weak point, the visibility problem is about content being citable, not about brand awareness.

Average mention position

Position is reported on a "lower is better" basis. In the product example, a brand at 2.8 sits mid-pack, behind a competitor at 2.1 and ahead of others at 3.4–4.7. Position tells you where in the answer you appear, which matters because early mentions tend to carry more weight in a short AI response.

Segmentation: platform and topic

A single blended score hides where you're actually winning or losing. Semji splits visibility two ways:

By AI platform. The product view shows a per-platform visibility score over the last 90 days:

• ChatGPT: 71
• Google AI Overviews: 58
• Perplexity: 54
• Gemini: 49
• Claude: 38
• DeepSeek: 22

This is the most actionable cut for most teams. A brand can be strong on ChatGPT and nearly absent on Gemini, which points to different content and source-selection behavior on each platform rather than a single "AI visibility" problem.

By topic. A topic filter lets you compare brands within a subject area, with score, mention rate, citation rate, and average position side by side. In Semji's illustrative table, five brands are ranked on the same topic, each with its own mention rate and position — so you can see not just your own trend but your standing relative to named competitors.

Tracking change over time

Every metric is reported with a delta, and the view is built around a time window (the example uses 90 days, updated weekly). The trend lines cover visibility score, mention rate, and citation rate together, with a market-average line for context. This is what turns a snapshot into a decision: a score of 62 means little on its own, but "62, +8 points this quarter, second of five tracked brands" tells you whether your content work is moving the needle.

What to check before relying on these numbers

• The example data is illustrative. Semji labels the brand comparison table as demonstration data, with brand names used for comparison only. Treat the specific figures as a format reference, not benchmarks.
• The metric set is defined by tracked prompts. Mention rate and citation rate are shares of the prompts you track, so the prompt set determines what the percentages mean. A narrow prompt set will produce confident-looking numbers that don't generalize.
• Platform coverage is a list, not a guarantee. The platforms shown are ChatGPT, Google AI Overviews, Perplexity, Gemini, Claude, and DeepSeek; confirm current coverage for your account rather than assuming all are included.

If your goal is to know whether AI systems cite you — and whether that's improving by platform and topic — these four metrics plus the two segmentations cover it. If you only need a directional read, the AI Visibility Score and its trend are enough; if you're diagnosing why visibility is low, the mention-vs-citation split and the per-platform breakdown are where the answer usually is.