AI recommendation monitoring and GEO optimization

AI Readiness Benchmark 2026: protocol before leaderboard.

This page publishes the study design. It does not claim benchmark results that have not yet been collected, reviewed and calibrated.

Research objective

Estimate the distribution of controllable AI readiness capabilities and observed AI representation outcomes across enterprises, while testing which capability measures remain reliable across sectors, locales and model changes.

Planned strata

Sector

B2B technology, professional services, finance, healthcare, manufacturing and consumer services, reported separately where evidence rules differ.

Market

English and Simplified Chinese tasks with country effects recorded rather than silently pooled.

Enterprise size

Size bands based on public operating scale, never inferred from website polish.

Evidence maturity

From fragmented facts to governed, versioned evidence systems.

Representation

Repeated task samples across multiple engines, dates and retrieval modes.

Agent surface

Presence and governance of documented actions, APIs and machine-readable interfaces.

Calibration design

  1. PilotTest rubrics on a small set with known positive and negative controls.
  2. Dual reviewMeasure inter-rater agreement and adjudicate material claims.
  3. StabilityRepeat samples to estimate within-site and within-model variance.
  4. External validityTest whether capability measures predict later observed outcomes.
  5. Version releaseFreeze the method, publish limitations and preserve raw lineage.

Participation and publication

Participation will require permission to test public properties, a fact owner for adjudication and agreement on anonymization. Commercial participation will never purchase a better score. Conflicts, sponsorship and exclusions will be disclosed with the release.

Status: protocol design and pilot recruitment. Benchmark values are not yet available.