Build a useful prompt cohort
Start with real buyer questions: category discovery, comparisons, alternatives, problems, product requirements, pricing, and regional intent. Keep benchmark prompts stable when measuring change.
Documentation
Guides for setting up prompt sets, running checks, and verifying evidence. Full docs are expanding ahead of launch.
- Getting started with prompt cohorts
- Running checks across five engines
- Regions and language targets
- Reading visibility and share of voice
- Inspecting raw answers and citations
- Exports and scheduling
Core workflow
From buyer question to auditable observation
Wolf visibility is built around defined checks rather than a shared black-box index. One check combines a prompt, an engine, and a regional or language target.
Choose engines and targets
Select the AI surfaces and supported regional or language targets that match the market. Wolf previews the required credits before execution.
Read the evidence
Review the raw answer together with detected brands, relative position, citations, and exact source URLs. A score without its evidence can hide important context.
Compare responsibly
Compare like with like: the same prompts, compatible engines, the same target scope, and successful responses. Record changes without treating correlation as causation.
Important operating notes
AI answers can vary between runs because engines change their models, retrieval systems, available sources, and product behavior. Location, language, time, and account context may also influence a response. Wolf preserves the metadata and raw evidence for a defined observation; it does not claim that one response is a fixed rank or represents every user.
A failed engine execution is not the same as a brand omission. Failed checks are recorded separately, excluded from successful-response metrics, and not charged. When an engine exposes citations or grounding sources, Wolf stores those URLs; when it does not, Wolf does not fabricate a citation layer.
Read the full methodology for metric definitions, or use the pricing calculator to estimate a prompt cohort before joining the launch waitlist.