GEO metrics vs SEO metrics: what maps and what doesn't
SEO dashboards are built on position, impressions and CTR. None of those exist inside a generated answer. Here’s the honest translation.
The mapping
| SEO metric | GEO equivalent | Why it maps |
|---|---|---|
| Position | Answer rank in entities[] | Order still correlates with being chosen |
| Impressions | Prompt coverage — how many monitored prompts return an answer naming you | Both measure presence |
| CTR | Referral traffic per cited URL | The only click-through signal that survives |
| Share of voice | Citation share of sources[] | Fraction of evidence you own |
| SERP features | Answer blocks (shoppingCards[], places[], ads[]) | Structured units, different names |
What doesn’t map
- Rank is a coordinate; a mention is a sentence. An answer can name you positively, neutrally or as the cautionary example. Sentiment has no SEO analog — you need the actual text.
- Positions are stable within a day; answers resample. Variance is a first-class property in GEO, which is why sample sizes and cadence matter.
- One SERP per keyword; many engines per prompt. ChatGPT, Perplexity and Gemini are separate markets — report per engine or the numbers lie.
- The query layer is hidden in SEO, visible in GEO.
searchQueries[]exposes the fan-out — searches the engine ran that you’d never have keyword-tracked.
The two-layer dashboard
The pragmatic setup keeps both pipelines: Google Search positions for the retrieval layer (fan-out queries still hit the index), monitor endpoints for the answer layer. The migration map walks the whole port, and share of voice defines the formulas.