Product evidence and controls
GEO and AI search observations
Generative Engine Optimization starts with observable responses and accessible website evidence. GrekSignal keeps individual AI samples separate from broad claims about search visibility.
01
Define a repeatable question
Save the question and configuration you want to observe. A manual run requires an official AI provider, an explicit model and both runtime and plan allowance. The resulting record identifies the provider, retrieval time and observation context; failed attempts also consume the reserved budget.
02
Inspect evidence and citations
Review brand mention annotations and citations actually returned by the provider. A saved response is a sample for that prompt, not evidence of a market-wide ranking. Missing citations and unavailable providers are shown directly instead of being converted into a score.
03
Review public AI readiness
A bounded site audit can inspect schema, llms.txt and public crawler directives. These are technical observations about the sampled pages. A crawler allowance or llms.txt file does not guarantee that an AI service will index, cite or recommend the business.
Practical questions
Is there a universal AI visibility percentage?
No. Saved response samples do not represent every user, model, location or prompt.
Can I begin without an AI key?
You can organize prompts and inspect public technical evidence. Actual provider observations require configuration and finite allowance.
Start with observable evidence
Run the bounded public scan on the homepage, then save a project when you are ready. Connected measurements need real authorization and available quota.