# GEO Score Specification — Changelog

All notable changes to the GEO Score Specification are documented here.

Versioning follows [Semantic Versioning](https://semver.org/):
- **Patch (1.0.x)**: Clarifications, typo fixes — no methodology changes
- **Minor (1.x.0)**: New signals added, engine biases updated — backward compatible
- **Major (x.0.0)**: Breaking changes to scoring formula — requires migration guide

---

## [1.0.0] — 2026-04-07

### Published

First public release of the GEO Score Standard.

**Signals:**
- 9 scoring signals defined (S1–S9), total max 100 points
- Signal weight table with mathematical formulas for each signal

**Engines:**
- Per-engine proxy scores for 5 AI engines: ChatGPT, Perplexity, Claude, Gemini, Grok
- Each engine has distinct signal weightings based on observed citation behavior

**Verticals:**
- 10 vertical categories: Cybersecurity, AI Tools, SaaS, FinTech, DevTools, LegalTech, HealthTech, Ecommerce, MarTech, HRTech
- Multi-signal vertical classification algorithm (6 evidence layers)
- Vertical-relative scoring and ranking

**Infrastructure:**
- Public specification page at traffi.app/geo-standard
- Version API endpoint at `/api/geo-standard/version`
- CC BY 4.0 licensing
- GEO Standard Working Group governance structure

**Validation:**
- Baseline correlation with sampled AI citations: r = 0.73 (Pearson) on 1,000-domain validation set (Q1 2026)

---

## [Unreleased] — Target Q3 2026 (v1.1.0)

### Planned

- **S10: Content freshness decay** — date-based signal penalizing stale content
- **llms.txt presence signal** — bonus for domains serving a well-structured llms.txt
- **Grok recency weight refinement** — improved modeling of xAI's real-time data preference
- **Answer-format match rate** — measure how closely content structure matches common AI response formats