Learn how to read prompt monitoring data, including brand mentions, sentiment, citation sources, and Share of Voice.
Last updated
Prompt monitoring captures what AI platforms say about your brand when users ask relevant questions. This page explains how to read each metric in your monitoring dashboard.
For setup instructions, see Set up prompt monitoring.
Each monitored prompt displays aggregated metrics across all AI platforms.
Click into any prompt to see the full AI response text from each platform, with your brand mentions highlighted.
Vizzybl tracks responses from eight AI platforms. Each platform has different training data and response patterns, so your visibility varies across them.
| Platform | Type | What it tells you |
|---|---|---|
| ChatGPT | Conversational AI | How OpenAI models represent your brand |
| Gemini | Conversational AI | How Google models position you |
| Claude | Conversational AI | How Anthropic models describe your brand |
| Perplexity | AI search engine | Search-focused responses with source citations |
| Grok | Conversational AI | How X/Twitter-integrated models reference you |
| Microsoft Copilot | Conversational AI | How Microsoft-surfaced answers describe you |
| Google AI Overviews | Search-integrated AI | AI summaries shown directly in Google Search |
| Google AI Mode | Search-integrated AI | Google's conversational search surface |
Per-platform metrics include brand mention count, domain citations, total citations, and competitors mentioned. Use the platform breakdown to identify where you are strong and where you need to improve.
The brand sentiment score ranges from -100 (entirely negative) to +100 (entirely positive). Each individual mention is scored for sentiment and then aggregated across all platforms and prompts.
Your dashboard also shows descriptors — the key phrases AI platforms use when talking about your brand (e.g., "industry leader," "affordable alternative," "limited documentation"). Descriptors help you understand how AI models characterize your brand beyond a numeric score.
When AI platforms cite sources in their responses, Vizzybl classifies each citation into one of five types.
| Type | What it means | Example |
|---|---|---|
| UGC | User-generated content | Reddit threads, forum posts, reviews |
| Editorial | Professional journalism | News articles, expert publications |
| Corporate | Your own content | Pages from your domain |
| Competitor | Competitor content | Pages from competitor domains |
| Other | Reference or utility content | Government sites, Wikipedia, legal resources |
A healthy citation profile includes a mix of corporate and editorial sources. A high share of competitor citations signals that AI platforms prefer competitor content for your target queries.
When you open an answer from Gemini, Google AI Overviews or Google AI Mode, each cited link shows the passage the engine quoted from that page underneath it. For Gemini that is how the passage opens and the words it ends with. For Google it is the snippet shown beside the source. Other engines do not say which part of a page they used, so their links show no passage.
To see the same passages ranked across all your prompts, open the Cited passages panel on the dashboard's Citations tab. See Coverage & Citation Analysis.
On shopping questions ("best … for …", "which … should I buy"), Google AI Mode often puts product cards inside its answer, each with its own reason for the pick ("Best for: …"). On Ultra and Enterprise plans, an opened Google AI Mode answer lists these under the answer text as Shopping picks Google showed: the product, its price, rating and seller, Google's reason, and a Yours or Competitor tag when Vizzybl recognises it from your brand, product catalog or competitor list.
Shopping picks are not ads, and their links go to Google Shopping, so they never count as citations. If your workspace also counts a pick as a mention of the brand or product it names, the strip says so.
To see every product Google picked across your prompts, open Agentic Optimization → Shopping Journey. See Agentic Optimization.
The competitor section shows which brands appear in the same AI responses as yours, with per-competitor mention counts and citation counts. This reveals who AI platforms consider your direct alternatives.
If you haven't configured competitors yet, see Identify your competitors.
Share of Voice measures your brand's presence across all monitored prompts.
(mentionCount / totalChecks) × 100A score of 50 means half of all brand-or-competitor mentions across your monitored prompts are yours. The metric that means "your brand appears in half of all AI responses" is Mention Rate, which the dashboard shows as a separate KPI card. The per-platform breakdown shows which AI engines mention you most and least.
The Prompt Bias Index measures how well your brand appears in discovery prompts compared to branded prompts.
(Discovery Mention Rate / Branded Mention Rate) × 100Discovery prompts are unbranded queries like "best consulting firms." Branded prompts include your brand name. A score of 100 means AI platforms mention you in discovery contexts as often as branded ones.
| Zone | Score | What to do |
|---|---|---|
| The Ghost | 0–20 | AI platforms rarely mention you organically. Invest in discovery-stage content targeting category queries. |
| The Echo Chamber | 21–40 | You appear mostly in branded searches. Create content that answers unbranded, category-level questions. |
| The Challenger | 41–70 | Growing organic visibility. Continue building content authority for unbranded queries. |
| The Authority | 71–110 | Balanced presence across branded and discovery prompts. Protect your position. |
| The Commodity | 111–200 | You appear in unbranded queries more than branded — branded recall is weakening. Audit branded prompts for competitor hijacking. |
Vizzybl generates three types of prompts from your keywords.
Tracking all three types gives you a complete picture of your AI visibility across the user journey.
Some prompts display a partial results indicator. This means one or more AI platforms failed to return a response for that prompt on its latest monitoring run — due to a timeout, rate limit, or temporary platform issue.
Only the latest run counts: a failure more than 36 hours old, or from a platform your plan no longer includes, does not keep the indicator on. If no platform answered on a run, the prompt also shows partial results rather than an error, so it stays monitored and is tried again on the next run.
Partial prompts still contribute to your metrics based on the platforms that did respond. Vizzybl retries failed platforms automatically in subsequent monitoring rounds, and the retry button only offers the platforms your plan includes.