Prompt corpus design
1,000 to 1,200 prompts built around your category, buyer intent signals, competitor mentions, and the specific question types that drive decisions in your market.
Model Behavior
How every major AI engine reasons about your brand right now.
We run 1,000 to 1,200 structured prompts across up to 9 AI engines and return the exact citation patterns, confidence markers, and recommendation logic for your brand and every direct competitor.
Get your answer intelligence report
We test your brand across AI engines before the call. Usually 20 minutes.
Method
How we build your answer intelligence map
Every corpus is built from scratch for your brand, your category, and your competitive set. No shared prompt banks.
Cross-engine testing
Each prompt is run across up to 9 engines — ChatGPT, Claude, Perplexity, Gemini, Copilot, Grok, Meta AI, AI Overviews — capturing the full response, citation URLs, and position data.
Citation extraction
Every brand mention, citation URL, recommendation position, confidence qualifier, and competitor co-occurrence is logged and structured for analysis.
Pattern analysis and briefing
We synthesise the corpus into a structured briefing: where you appear, where competitors win, what evidence each engine uses, and what to build next.
What the corpus reveals
Six layers of answer intelligence
Citation frequency maps
Which sources each engine retrieves for your category, how often, and in what context. Not assumption — measured output per prompt.
Confidence attenuation
Where the engine hedges, qualifies, or avoids recommending your brand. Hedges are signals: they tell you exactly what evidence is missing.
Recommendation trigger language
The exact phrasing, intent patterns, and query structures that cause each engine to recommend your brand — and the ones that consistently surface competitors instead.
Competitive position trace
For each prompt type, who gets recommended, in what position, with what justification. A ranked view of your share versus every competitor in the corpus.
Engine-to-engine divergence
Where ChatGPT and Perplexity disagree on your brand. Divergence points to missing evidence in one pipeline — parametric or retrieval — and tells you which to fix first.
Retrieval source inventory
The specific URLs, domains, and content types each engine cites for your category. The foundation for every placement and distribution decision that follows.
What the briefing includes
A structured document you can hand to any agency, content team, or PR firm and immediately act on.
Answer share report
Your citation percentage across all 9 engines on every prompt category. The number that all future work moves.
Competitor gap map
A structured view of where competitors appear and you do not — broken down by engine, prompt type, and evidence type.
Evidence priority list
Ranked actions: which evidence to build first, which sources to target, and which engine to focus on for fastest measurable movement.
12-week delta test
The same prompt corpus re-run after 12 weeks of work. Movement is measured against your own baseline — not an industry average.
Outcomes
What you walk away with
Engines tested
ChatGPT, Claude, Perplexity, Gemini, Copilot, Grok, Meta AI, AI Overviews, and more depending on your market.
Prompts per run
Not a sample. A full corpus across every intent type, competitor co-mention, and category signal relevant to your brand.
Baseline set
Your answer share number is established before anything is changed. Every result is measured against this, not claims.
Priority actions
Not a list of ideas. A ranked, engine-specific set of evidence gaps and the exact sources to place in to close them.