Platform/Pillar 5
Private beta

Optimization

Turn Content Genome findings into specific, typed recommendations — then deploy them through an approval-gated, verifiable system that leaves an immutable record.

From content understanding to action

The Optimization Engine is where observation becomes intervention. Once the Content Genome has characterized a client's site — identifying what's present, what's structured, and what's missing — the engine applies a set of deterministic rules to generate typed recommendations. Each recommendation proposes a specific, exact DOM-level change: not 'improve your structured data' but 'add LocalBusiness schema to your homepage with these specific fields at this location.'

What the engine can currently recommend

Six recommendation types are live in the current Optimization Engine: missing LocalBusiness schema, missing FAQ block, thin content, missing meta description, multiple H1s, and missing service area block. Each rule is deterministic — the same content genome produces the same set of recommendations on each run. New recommendation types will be added as the engine grows; all capability additions are logged in the changelog.

The approval gate: how changes get deployed

No recommendation deploys automatically. Every intervention passes through a risk-aware approval workflow: the recommendation is packaged and signed, preflight checks run against the live site, and the agency reviews and approves before anything changes. Once approved, the intervention is deployed via the Standard Script — a lightweight browser tag that applies constrained, typed-DSL primitives to the DOM at runtime. After deployment, GeoScript independently verifies that the change landed as intended.

Standard Script: how GeoScript touches a site

The Standard Script is a fail-open browser tag installed on a client's site. It fetches a signed deployment manifest from GeoScript and applies typed primitive operations — the DSL explicitly rejects inline scripts, event handlers, and javascript: URLs. Changes are bounded, auditable, and rollbackable. If the manifest fetch fails, the page loads normally with no intervention applied.

How this connects forward and back

Optimization lives in the middle of the loop. It consumes Website Intelligence (the Content Genome) on one side, and its outputs flow into Reporting and outcome measurement on the other. After each deployment, GeoScript re-crawls the site, re-runs AI visibility observations, and attributes changes in visibility to specific interventions. The loop is closed through Reporting.

What Optimization is not

Explicit limitations (§127)

  • GeoScript does not automatically deploy any change — every intervention requires agency approval before it reaches a client's site.
  • GeoScript does not guarantee that a deployed change will increase AI visibility — it proposes changes based on structured content evidence; outcomes are measured separately.
  • The Content Studio (human-review surface for editing recommendations before deployment) is planned but not yet available.
  • GeoScript does not modify server-side content, CMS databases, or backend systems — Standard Script operates at the browser DOM layer only.
  • GeoScript does not claim causal attribution for visibility changes — it measures before-and-after observations and reports the difference; causal inference is explicitly noted as imperfect.

Related

Continue through the loop

Optimization is pillar 5 of 7.

Other pillars