Methodology

This leaderboard ranks individuals and agencies by how prominently leading generative models name them when asked "who are the top AEO / GEO experts".

1. Prompt battery

We run a fixed set of 0 prompts every run. Editing prompts starts a new season so historical scores stay comparable.

2. Sampling per model

Each prompt is run 3× per model on default temperature to smooth out single-shot randomness. Currently active models: ChatGPT, Gemini.

3. Mention extraction

A small extractor model parses each answer into structured JSON: name, type (person / agency), and rank in the list (if numbered). Names are normalized against an alias map so variants like "Andrew Ng" and "Dr. Andrew Ng" collapse.

4. Position scoring

Each mention earns a position score using DCG-style weighting: 1 / log₂(rank + 1). Mentions in prose (no ranked list) score 0.5. Per-model scores are normalized to 0–100.

5. Composite (Overall)

The Overall score is the equal-weighted average of the normalized per-model scores across all active models.

6. Refresh cadence

The pipeline runs weekly. The first public run will appear shortly.

7. Historical archive

Every weekly run is preserved in the run history. This archive lets you browse past leaderboards, compare how rankings and mention counts have shifted over time, and inspect the full raw model responses that produced each run's scores.