Prompt Monitoring
Author, schedule, and version the prompts GeoScript runs across AI engines to build a longitudinal visibility signal.
Private beta — available to agency partners
This capability is live and working for agencies in GeoScript's private beta program. Learn about agency access →
What Prompt Monitoring does
Prompt Monitoring is the system that defines, versions, and schedules the queries GeoScript submits to AI engines. Every client gets a deterministic prompt universe — approximately 1,100 prompts across five categories — generated from their business profile. These prompts are the lens through which AI visibility is measured. The same prompts run each cycle so results are comparable. Prompts are versioned, never deleted, and retired rather than removed — preserving the historical record.
How it works
Deterministic prompt universe generation
GeoScript generates a client's prompt universe from their business profile inputs — category, geography, services, and other signals. The generator is deterministic: the same inputs produce the same 1,100+ prompts every time. Prompts are identified by SHA-256 hashes of their content, making it possible to detect when a prompt has changed and what version it was at any point in time.
Five prompt categories
The prompt universe is organized into five categories: service discovery ("who offers X near Y"), comparison ("best X in Y vs Z"), local (geography-specific queries), reputation ("is X reliable / trusted"), and category-level (industry or vertical queries). Coverage across categories means GeoScript captures a client's AI visibility profile across the full range of query types a real user might ask.
Idempotent upsert and pagination
When a prompt universe is generated or refreshed, GeoScript uses an idempotent upsert — if the prompt already exists in the system, it is updated in place rather than duplicated. Prompts are never deleted; they are retired with a timestamp when no longer active. The pagination system supports more than 1,000-row query sets, proven in smoke tests.
Scheduling and execution pipeline
The QueryDefinition table stores each prompt's definition, schedule, and version history. The execution pipeline picks up due prompts, submits them to the configured AI engines, and records the raw responses as immutable observations. The pipeline is real end-to-end — 22/22 smoke tests passing.
What Prompt Monitoring does not do
Clear boundaries increase trust. These limitations are documented explicitly, not buried in footnotes.
Prompt Monitoring does not allow agencies to submit arbitrary prompts to AI engines outside the structured prompt universe — queries run through GeoScript's pipeline, not a free-form API proxy.
The prompt universe is not manually curated for each client — it is generated deterministically from the client's profile. Agencies can influence inputs but not manually author individual prompts in the current version.
Prompts are not guaranteed to remain stable across all future GeoScript versions — formula updates produce new versioned prompt sets.
GeoScript does not control or influence how AI engines respond to prompts — only what prompts are submitted.
Related
Other features
GeoScript is in private beta
Agency partners get access to the full capability set — acquisition, monitoring, website intelligence, optimization, deployment, and reporting — for $299/month per active client.