Deldai documentation
AI Visibility monitoring
Use repeatable query sets to inspect how a configured AI provider responds to questions relevant to your brand or subject.
How a run works
- Open AI Visibility in the dashboard and allow the query sets and provider status to load.
- Select a query set. A run cannot start until a query set is selected.
- Choose a provider from the providers available in your environment.
- Choose a run mode supported by that provider.
- Start the run and wait for its status and measurements to update.
- Open the run details to inspect the available measurements and related comparison or alert information.
Understand the run modes
Model response
Captures responses from the selected provider in the model-response mode. It does not, by itself, establish how a live search results page ranks a website.
Web-grounded
Uses the provider's supported web-grounded mode when that mode is available and configured. The availability and behavior depend on the provider integration.
Read run status carefully
- Pending: the run has not finished processing.
- Running: processing is in progress.
- Completed: processing reached a completed state; review the counts and measurements to understand coverage.
- Failed: the run did not complete successfully; review the displayed error or available details.
A completed run can still contain failed individual queries. Review completed and failed counts rather than assuming every query succeeded.
A run is not necessarily a schedule
Starting a run executes a monitoring-purpose run for the selected query set. Do not assume it will repeat automatically. Use a scheduling control only if that control is available in your workspace, and confirm its saved schedule and status in the product.
Interpret findings with context
- Use consistent query sets when comparing runs.
- Record the provider and mode because they affect the response being measured.
- Read the returned answer and source information where available; do not infer a citation or recommendation that is not present in the result.
- One run is a snapshot, not proof of universal or permanent AI visibility.