Getting started with M2 Visibility
From a business question to the first measurement cycle: how to structure brands, competitors, prompts, models and decision criteria without turning the platform into another dashboard.
How to interpret this document
This content describes technical and methodological behavior that is implemented or explicitly planned in the product. When a control depends on configuration, a provider, a secret, a contract or legal approval, that dependency must remain visible.
The problem the platform solves
AI systems increasingly mediate discovery, comparison and recommendation. The question is no longer only where your brand ranks in search, but how it is described, recommended, compared and supported by evidence inside AI-generated answers. M2 Visibility turns that phenomenon into a measurable operating discipline.
Start from the decision
Before creating prompts, define what the organization needs to decide. Examples include understanding why a competitor is more frequently recommended, identifying which external sources support a narrative, testing purchase-intent visibility or determining whether a content action changed a comparable result.
Configure brand and context
Set the monitored brand, aliases, competitors, market, country, language and organizational scope. These elements become part of the measurement context and help prevent incompatible comparisons.
Treat prompts as strategic assets
Prompts should represent real market questions and can be classified by intent, funnel stage, persona, business importance and strategic weight. prompt_key provides a stable identity for comparison and statistics.
Choose models and methodology
Select supported models and the appropriate measurement mode. Natural measures spontaneous recall. Grounded uses current web evidence. Competitive frames market comparison. Reputation explores public perception. Purchase Intent moves closer to an actual buying decision.
Choose repetitions
Generative answers vary. Repetitions help observe stability for each prompt+model unit and improve statistical quality. A single answer should be treated as an example, not durable evidence.
Run and interpret
A MeasurementRun creates MeasurementJobs for each prompt, model and repetition combination. Completed responses feed AVS-2.0, STATS-1.0, Source Intelligence and other analyses allowed by entitlements.
Close the loop
The intended cycle is Measure → Understand → Act → Prove. Recommendations, alerts and ownership convert evidence into action. A later compatible run tests whether the signal really changed.