Reading your share-of-voice dashboard

The four panels on the main dashboard

• Overall SoV — your weighted share-of-voice across every tracked prompt, every engine. The headline number.
• Per-engine breakdown — same number split by ChatGPT / Perplexity / Gemini / Claude. Often dramatically different.
• Top movers — prompts where your SoV moved most (up or down) day-over-day.
• Competitor diff — your SoV vs your pinned competitors on the same prompts, per engine.

What's noise vs what's a real shift

Day-to-day variance of ±3-5 percentage points is normal — LLM responses aren't deterministic, and small fluctuations in retrieval mean tomorrow's answer can differ from today's. Trust trends over single days. A 5-point drop on one prompt for one engine on one day is noise. A 5-point drop sustained over a week, especially across multiple engines, is a real shift.

Acting on a drop

1. 1Open the affected prompt and read the actual LLM response. Who's being recommended instead of you?
2. 2If a competitor: check their backlinks and content changes in the last 30 days (Backlinks tab → Competitor view). Likely they shipped or earned something.
3. 3If a generic answer ("users typically choose..."): the LLM may have downgraded its specificity — recover with fresh content that establishes you as the canonical answer for that prompt.
4. 4If you've been removed in favor of a directory or aggregator: focus on getting mentioned in those aggregators, or on creating richer content that the LLM would prefer to cite directly.