How We Test AI Companions
Attraction gets the click. Testing earns the recommendation.
We deliberately separate personal fit, product quality and commercial performance.
Match Score
Based on the appearance, personality, archetype, visual style and feature preferences the visitor selected. Affiliate commission is excluded.
Provider Test Score
Standardized testing covers conversation quality, repetition, initiative, roleplay, personality stability, memory at multiple time horizons, images, video, voice, value and privacy.
Real Cost
We record the advertised subscription separately from estimated light, normal and heavy-media use so token and media spending does not disappear behind a headline price.
Evidence
Provider promises and source-page facts are stored separately from firsthand benchmark observations. AI-assisted extraction is reviewed before it becomes published data.
Long-term memory
Memory can be logged at 10 messages, one hour, one day, seven days and thirty days. An untested field stays unscored instead of being guessed.
Commercial data
Clicks, conversions and revenue are visible in the admin. EPC can break a near-tie between similarly suitable offers, but it cannot change public Match or benchmark scores.