GEO Platform · Generative Engine Optimization
WICHI
The SEO of the AI search era. Measures and improves how brands surface across ChatGPT, Gemini, and Perplexity.
Live at wichi.app · paid plans active · financial-compliance add-on in development · 2026–
What it is
WICHI is a GEO platform — Generative Engine Optimization — built for the discovery surface that AI search created. It tracks how a brand appears across the three engines that now answer most discovery queries (ChatGPT, Gemini, Perplexity) and reduces noisy, single-prompt anecdotes into a stable, comparable signal: a 4-layer GEO Score over a 9-bucket query framework with LLM-Judge evaluation.
GEO Score
43.1
/ 100
Mention Rate
71.8%
in AI responses
Avg SOV
43.0%
vs competitors
Hallucination Risk
26.9%
⚠ error rate
Group Performance
| Group | GEO | Mention Rate | SOV |
|---|---|---|---|
| Owned | 63.7 | 94.0% | 88.0% |
| Battleground | 30.4 | 55.0% | 25.0% |
| Competitive | 55.9 | 78.0% | 55.0% |
GEO Score by Bucket
Queries (Sample)
| Bkt | Query | Mention | SOV | GEO |
|---|---|---|---|---|
| A | Does the Samsung &Mileage Card have any event promotions? | 100% | 73.2 | |
| B | Can I access airport lounges with the Samsung &Mileage Card? | 100% | 66.5 | |
| C | Recommend credit cards with good mileage rewards, especially for international use. | 0% | 18 | |
| F | Compare Samsung &Mileage Card vs KB Kookmin Card for international mileage. | 28% | 56.2 |
Why it matters
Buyers increasingly ask ChatGPT before Google. The measurement layer SEO was built on no longer covers the actual discovery surface — and a single prompt is not a measurement. WICHI is the measurement layer for the post-Google era, evidence-grounded enough to be auditable and stable enough to drive decisions.
How it works
WICHI runs a multi-agent pipeline of specialized LLM roles. Lightweight upfront work — like expanding a brand into dozens of intent-diverse queries across 9 buckets — uses mini-class models. The heavier collection-and-judge work — fanning queries out to ChatGPT, Gemini, and Perplexity, then scoring each response on visibility, prominence, sentiment, and hallucination — runs on larger production engines. The 4-layer GEO Score is a weighted aggregate designed for stability across reruns rather than single-prompt noise.
Status
Live at wichi.app. Currently expanding with a financial-compliance add-on as the next vertical layer.
Related reading on Minbook
- WICHI Build Log — full series, hackathon rejection through production GEO SaaS.
- Multi-Engine Architecture — 3 AI Search Engines in Parallel — how the multi-agent evaluation pipeline is wired.
Links
- wichi.app
- Demo and details on request — request via email