Stop guessing which prompts to monitor.
We'll handle that for you.
Powered by CAVIS, the engine behind CitationLab's AI monitoring
Generate your AI conversations for free
Enter your website and we'll build the actual multi-turn conversations your customers have with AI about your industry.
Why single prompts tell you nothing
Most AI visibility tools send a handful of isolated questions to ChatGPT and report whether your brand was mentioned. That approach has a fundamental flaw: it doesn't reflect how real people actually use AI search.
Real customers don't ask a single question and stop. They have conversations. They start with a problem, explore solutions, compare options, and narrow down to a provider. Each step in that journey changes what the AI recommends. A brand that appears when someone asks "what is X?" may vanish entirely when they ask "who should I choose?" And it's the second question that drives revenue.
Single-prompt monitoring is a snapshot of a moving target. CAVIS captures the entire film.
What a CAVIS simulation captures
Each simulation follows a customer through their complete decision journey. At every stage, CAVIS measures what matters.
Problem awareness
The customer describes a symptom or need. AI responds in educational mode. Brands are typically absent here and that's expected.
Solution exploration
The customer asks what options exist. AI begins naming categories and approaches. The first brand mentions often appear here.
Product comparison
The customer evaluates specific solutions. AI structures competitive comparisons. This is where positioning is won or lost.
Vendor selection
The customer asks for recommendations. AI names specific providers. This is the highest-value visibility moment in the conversation.
Commitment
The customer asks operational questions about a named provider. Maximum brand exposure and direct entity response.
What you get
CAVIS produces structured, actionable intelligence. Not vanity metrics.
Composite Visibility Score
A single 0-100 score that combines when your brand first appears, how prominently, whether it persists, and how it performs against competitors. Updated with every monitoring cycle.
Cross-model comparison
See exactly how your visibility differs across ChatGPT, Gemini, Perplexity and Google AI Overview. Identify which models require different content strategies.
Competitive displacement
Understand whether you're above or below your competitive set in AI recommendations. Visibility is zero-sum: if AI recommends three providers and you're not one of them, a competitor has taken your slot.
Reliability measurement
Know whether your visibility is stable or contested. High variance means small content changes could shift your position significantly. That's both a risk and an opportunity.
Why CAVIS outperforms prompt-based monitoring
| CAVIS | Prompt-based tools | |
|---|---|---|
| Simulation approach | Complete multi-turn conversations following real customer journeys | Isolated single prompts with no conversational context |
| Intent coverage | Full funnel from problem awareness to purchase decision | Random or keyword-based questions |
| Scoring model | Intent-weighted, position-aware, sentiment-adjusted composite score | Binary mentioned/not-mentioned or simple frequency counts |
| Model coverage | Simultaneous cross-model analysis with agreement scoring | Usually single-model or sequential testing |
| Statistical validity | Multiple runs with confidence intervals and reliability metrics | Single-run snapshots with unknown variance |
| Academic foundation | Grounded in IR theory, NLP research, and information theory | Ad-hoc methodology without theoretical basis |
Built on science, not guesswork
CAVIS integrates established research from information retrieval theory, conversational search, generative engine optimisation, intent classification, and information theory. Every component of the framework has a theoretical foundation and a measurable output.
Information Retrieval
Position-weighted scoring grounded in nDCG methodology (Jarvelin & Kekalainen, 2002)
Conversational Search
Multi-turn query structure and conversation-state modelling (Mo et al., 2025; Zamani et al., 2023)
Generative Engine Optimisation
Visibility metrics for generative AI responses (Aggarwal et al., KDD 2024)
Information Theory
Entropy-based measures for brand concentration and model reliability (Shannon, 1948; Farquhar et al., Nature 2024)
Questions about CAVIS
CAVIS is CitationLab's proprietary framework and a core part of our competitive advantage. We share the principles, outcomes, and academic foundations publicly, but the specific scoring models, formulas, and simulation architecture are confidential.
Most tools send individual prompts to AI models and check for mentions. CAVIS simulates complete multi-turn customer conversations that follow the natural purchase journey. This captures visibility dynamics that single-prompt approaches structurally cannot measure: when your brand first enters the conversation, whether it persists, how it performs at the critical vendor-selection stage, and how consistently each model recommends you.
CAVIS currently runs simulations across ChatGPT, Google Gemini, Perplexity and Google AI Overview. Each model has different retrieval mechanisms, training data, and citation behaviour. CAVIS measures these differences and produces a cross-model visibility profile.
Monitoring frequency depends on your plan. Standard monitoring runs regular simulation cycles to track visibility over time. Because AI models update continuously, regular monitoring is essential to detect both improvements and regressions.
CAVIS integrates and extends peer-reviewed research from multiple academic traditions, including information retrieval (Jarvelin & Kekalainen, 2002), conversational search (Zamani et al., 2023), generative engine optimisation (Aggarwal et al., KDD 2024), and information theory (Farquhar et al., Nature 2024). The framework itself is proprietary and not published, but its theoretical foundations are established science.