Skip to content

AI Visibility Tracking for Generative Engine Optimization

The ZSA GEO tracker (geo-tracker.py) is an internal telemetry tool in active development. It is intended to evaluate brand mentions and prompt-level Share of Voice (SOV) across AI model families without confusing generative engine optimisation with local or geographic rank tracking.

Development status: this is not currently a public self-service product. Model coverage, automation and evidence capture are being rebuilt and validated. The implementation notes below describe the present tracker configuration, not a promise that every model or reporting surface is continuously available.

The 13-Model Tracker Matrix

The current implementation is configured with a 13-model overlap matrix. Internal tiers compare Best (reasoning/flagship), Middle (workhorse/consumer) and Fast (lower-latency) model classes. These labels describe model roles, not user demographics.

OpenAI Ecosystem

  • GPT-5.5 Pro (Best)
  • GPT-5.4 Pro (Middle)
  • GPT-5.4 Mini (Fast)

Anthropic Ecosystem

  • Claude Opus 4.7 (Best)
  • Claude Sonnet 4.6 (Middle)
  • Claude Haiku 4.5 (Fast)
  • Claude Sonnet 5 (New flagship overlap)

Google Ecosystem

  • Gemini 3.1 Pro Preview (Best)
  • Gemini 3 Flash Preview (Middle)
  • Gemini 3.1 Flash-Lite Preview (Fast)

xAI Ecosystem

  • Grok 4.20 (Best)
  • Grok 4 (Middle)
  • Grok 4.1 Fast (Fast)

Current implementation and limits

  • Sends controlled test prompts to configured models through OpenRouter when the required credentials and endpoints are available.
  • Records binary brand/domain mention flags in citations.csv and calculates prompt-level Share of Voice from those observations.
  • Does not yet provide the transcript, citation-source, entity-resolution and consumer-interface evidence required for a production-grade visibility report.
  • Historical cron and leaderboard components exist in the repository, but continuous automation and a public leaderboard are not current public capabilities.

For the current client-facing diagnostic, see the AI Visibility Baseline. The full tracker page will be revisited after the measurement workflow and evidence contract are validated.