Prompts are the buyer questions your scans ask. Understanding how they work explains everything else in your report.
The default prompt set
Every scan starts from 29 buyer-question templates chosen for your industry and location. They are written the way real buyers ask — "best {category} in {city}", "which {category} is good for {use case}", "is {brand} trustworthy" — not as SEO keywords. Templates are grouped by tags (discovery, comparison, trust, local, …) so the report can show which kind of question you win or lose.
The channels
Each prompt is asked on 3 channels:
- Grok 4.5 (direct) — what the model answers from its own knowledge.
- Grok Composer 2.5 Fast (direct) — a second model for cross-model comparison.
- Grok 4.5 + Brave Search (grounded) — live Brave web results are fed to the model, approximating how retrieval-augmented assistants answer. This channel is also the main source of citations.
29 prompts × 3 channels ≈ 90 sampled answers per scan. Direct channels show what models believe; the grounded channel shows what the live web tells them. A gap between the two is itself a finding — e.g. strong grounded presence but zero direct-model presence means the models haven't internalized your brand yet.
We do not currently query ChatGPT, Gemini, Claude or Perplexity — see coverage.
Why templates beat ad-hoc questions
Asking the same questions every week is what makes trends meaningful. If the questions changed each scan, a moving mention rate could mean anything. Templates hold the measurement constant so changes reflect the world, not the questionnaire.
Customizing
With a monitoring dashboard you control the prompt set: add your own questions, edit or disable templates, and keep the scan focused on what matters to your business — see Add, edit and disable prompts. The next weekly scan asks exactly your enabled prompts.