Skip to main content
The GEO Score summarizes how strongly your brand appears across the AI engines and prompts Cleo tracks for a site. Use it to follow direction over time, then open the supporting metrics to understand why it moved.

​
The score components

The composite GEO Score weights these dimensions as follows: AI Presence Rate 20%, Grounding Frequency 20%, Recommendation Rate 25%, Share of Voice 20%, Share of Citation 10%, and Sentiment 5%. If a component is unavailable, Cleo recalculates the score over the available components. Answer Prominence and Citation Rank Index remain useful supporting analytics, but they are not inputs to the headline GEO Score. Citation Drift and Measurement Precision also sit beside the score because they describe stability and confidence rather than brand performance.

​
How Recommendation Rate is decided

There is no numeric threshold. For every answer that names or cites your brand, an AI model reads the answer and labels the brand one of three ways:
  • Recommended: the answer endorses you, suggests you, or presents you as a solution. For example, “Brand X is a great option for…”.
  • Mentioned: the answer names you without endorsing you. For example, “providers include X, Y and Z”.
  • Not mentioned: the brand does not appear.
The model judges mainly on the wording. It also considers whether you are named early and prominently or buried in a list, so a tenth-place “you may also consider” is less likely to count than a top pick. Recommendation Rate on the dashboard card is the share of answers that name or cite you and also recommend you. The label is yes or no, so a strong top pick and a weaker recommendation that still qualifies count the same. Your position among an answer’s sources is measured separately by Where AI lists you on the Citations sub-tab. See Citations and claims.
  • Compare weeks and months, not isolated runs.
  • Check the engine and prompt breakdown before assigning a cause.
  • Use the same tracked prompt wording if you want a continuous history.
  • Expect wider uncertainty when a site has little data or few tracked prompts.
  • Treat a sudden jump as a reason to inspect the evidence, not as proof of durable improvement.

​
What can move the score

The score may change when:
  • Your pages become easier to retrieve, understand, or cite.
  • Third-party sources begin mentioning or validating your brand.
  • A competitor gains or loses visibility.
  • An AI engine changes its answer or source selection.
  • Cleo adds new valid data for a tracked prompt.
Changing your prompt list is designed not to move the score on its own. A new prompt starts in Discovery, which is not scored, and counts only once it is promoted, arriving with its own measured result. A prompt you retire keeps counting at its last measured result for 12 months, so tidying the list does not cause a sudden drop. Moving a prompt between stages changes how often it runs, not how it is scored. See Prompt lifecycle. Score history starts on the day Cleo begins tracking the site. Cleo does not back-fill earlier data, and the chart continues as one series when you change prompts.
Use What moved your score? and per-engine views when available. They provide more useful context than the composite number alone.

​
Measurement precision

AI answers are not deterministic. Cleo reports sample size and confidence information so you can separate a meaningful change from ordinary variation. The confidence calculation uses the distinct tracked queries in the latest measurement snapshot; it does not combine query counts from earlier runs.
Do not promise a client a precise business outcome from a small score movement. Pair GEO trends with citation evidence, analytics, and completed work.