Methodology
How the GEO Score is calculated - full transparency
A number without visible methodology is just a promise. Here's exactly what goes into the GEO Score, what's left out, and why it matters so you can trust the result.
What the GEO Score measures
The GEO Score is an index from 0 to 100 that measures a brand's organic visibility in LLM responses - that is, what the model already knows about the brand from its training, without web research and without additional context injected into the prompt.
This is a deliberate methodological choice. Web search and RAG simulation (responses with injected website context) measure different things, and therefore appear separately in the gaps panel - mixing them in the same number would make the score difficult to interpret and easy to manipulate.
The three components of the calculation
1. Frequency of mention
How regularly the brand appears in responses to relevant prompts in your sector. A brand mentioned in 8 out of 10 tested prompts has a very different base score than one mentioned in 1 out of 10.
2. Relative position in the answer
Being the first brand mentioned in a response weighs more than appearing as a side note at the end of a list of six competitors. Position reflects how central the model considers the brand to answer the question.
3. Quote quality and tone
Not all mention is equal. An actively recommended brand ("the best option for X is...") scores differently than a brand just listed without comment, or mentioned in a neutral/comparative tone.
Calculation by engine, then aggregated
Each configured engine (ChatGPT, Grok, Claude, Perplexity) is evaluated separately, with the same set of prompts. The final GEO Score is the average of the scores obtained on each engine within the active plan.
This means that a brand can have a high GEO Score on ChatGPT and almost zero on Perplexity - and the aggregate number alone doesn't tell this story. That's why the detail per engine is always visible, not hidden behind the average.
Every point is verifiable
A score that cannot be audited is no different from an opinion. Therefore, each GEO Score value corresponds to real saved responses - the exact text returned by the model, by prompt and by engine, available for consultation.
If a brand disagrees with a score, it can see exactly which response generated that number, and not just rely on a “black box”.
What is not included in the GEO Score
- Real-time web search results - appear as a separate metric ("web search"), not mixed into the organic score.
- Mock responses with site context injected (RAG) - appear in the gaps panel for comparison, not in the main score.
- Traditional Google positioning (organic SERP, AI Overviews) - is a separate metric, available separately when the Google tracking add-on is active.
How to interpret the number
The GEO Score is not an absolute "site quality" number - it is a trend indicator about a brand's presence in the organic awareness of the models consulted. A low score does not necessarily mean a poorly designed website; It could just mean a brand whose online presence is too recent to have already been incorporated into the training of current models.
That's why the GEO Score is best read side by side with RAG-Readiness - the technical preparation of the website to be retrieved by AI systems. A site with high RAG-Readiness and low GEO Score has the right content, but it has not yet been "learned" by the models. It is a sign of opportunity, not failure.