Which AI engines we scan and why
The seven engines we cover today, why we chose them, and what we do with the results.
Which AI engines we scan and why
TL;DR — Seven engines: ChatGPT, Perplexity, Gemini, Grok, Copilot, Google AI Mode / AI Overviews, and DeepSeek. Together they cover the overwhelming majority of AI-answer traffic globally.
The seven engines
| Engine | Who | Why it's in scope |
|---|---|---|
| ChatGPT | OpenAI | Largest consumer AI surface, deep enterprise footprint |
| Perplexity | Perplexity AI | AI-first search with the cleanest citation model |
| Gemini | Consumer chat + tight Google-app integration | |
| Grok | xAI | Growing consumer + X-embedded surface |
| Copilot | Microsoft | Enterprise reach via Microsoft 365 + Bing |
| Google AI Mode / AI Overviews | AI-generated answers embedded above classic search | |
| DeepSeek | DeepSeek | Fast-growing globally, dominant in several markets |
Why not Claude in the visibility scan?
Claude is fantastic and we use it internally — but Anthropic doesn't currently expose the kind of public answer surface that gives GEO its measurable signal. When that changes, we'll add it and back-populate history from that day forward.
Why not just "the big three"?
Different buyer segments live on different engines. Perplexity indexes the most research-heavy professionals; Grok reaches an audience Google Search often misses; Copilot dominates certain enterprise workflows; DeepSeek is the default assistant in several regions.
Scoring against only ChatGPT would over-fit your strategy to one platform and one audience.
How we hit each engine
We use a mix of direct APIs (where available), official SDKs, and SERP-backed extraction (via Bright Data) for the engines that don't expose their answer generation via API. Every response we log is a real captured answer, not a simulated one.
The one-formula guarantee
Every engine is scored using the same formula: mentions ÷ responses × 100. That means your Perplexity score and your ChatGPT score are directly comparable — a 25% on both is genuinely the same amount of visibility.
We do not weight bigger engines more heavily in the roll-up. The overall score is the arithmetic mean across engines, treating each equally. That's opinionated — the reasoning is that a customer who lives on Perplexity shouldn't have their reality diluted by your ChatGPT numbers.
What we don't do
- No fake benchmark comparisons against competitor tools by name.
- No hidden "confidence intervals" that let us fudge numbers.
- No "estimated" mentions when we didn't capture an actual response.
Every point on your chart is a real captured AI answer, on a real prompt, on a specific day.