See how AI answers mention your brand, with the evidence attached
Group prompts by topic, run them repeatedly, and measure how often answers mention or cite you. Every number shows its sample size, failed samples and the exact answers behind it.
Experiments are labelled as experiments
Calling a model’s API is not the same as measuring what a consumer AI app shows a real user. VisiRelay records the surface, model, grounding mode, locale (only when guaranteed) and time of every sample, and keeps API experiments separate from observed search surfaces.
Mentions and citations are different facts
Brand mention rate counts answers that name a verified alias. Citation presence counts answers that link to your domain. Share of mentions uses a competitor set you define and can see.
- Repeated samples per prompt, with failures kept visible
- Answer text and returned citations stored as evidence
- Proposed new topics and aliases need human approval
What it doesn’t do (yet)
- We cannot see an AI system’s training sources; only returned or observed citations count as evidence.
- Sentiment and theme tags are inferences with a stored classifier version, not facts.