Skip to main content
Tracked prompts define what Cleo measures. A useful prompt represents a real question your audience asks and an outcome where your brand has a credible reason to appear.

​
Choose strong prompts

Good prompts are:
  • Specific enough to produce a useful answer
  • Commercially or strategically relevant
  • Written the way a customer would ask
  • Stable enough to track over time
  • Broad enough that more than one credible answer exists
Avoid prompts that contain the answer, ask only for your brand, or are so broad that the result has no decision value.

​
Manage prompts

Open GEO → Prompts to add prompts up to the site’s plan limit. You can add your own, and Cleo suggests prompts from the site and its configured industry. The two work together. What you can change depends on the prompt’s stage:
  • Discovery: edit the wording, or Drop the prompt. Discovery prompts are not scored, so neither changes your score.
  • In Play, Earned, Retired: the wording is locked. Changing the wording of a prompt that already has history would break its own trend line. To track a different question, add a new prompt.
  • In Play: Stop tracking becomes available once the prompt has shown no sign of your brand for 30 days. It is permanent.
  • Earned: Retire becomes available when the prompt is fully won (4 of 4). A retired prompt keeps counting in the score at its last measured result for 12 months.
Editing prompts never recomputes past readings. They stand as measured, and changes apply from the next run.

​
Research prompts before you track them

The Research stage, before Discovery, shows what people ask AI about a topic before you track anything. For each question it shows the AI search volume, its trend, and which brands AI names in the answers, and it ranks suggestions by volume. Research data covers ChatGPT and Google AI Overviews. A strong tracking prompt has real volume, a trend that is holding or rising, and a set of named brands you can realistically join.

​
Region

Prompts run for the country set on the site, and the default is the United States. The region shapes the answers: a mobile network prompt surfaces Airtel and Jio for a site set to India, and T-Mobile for a site set to the US. The region applies to every prompt on the site; a different region per prompt is not available.

​
Prompt lifecycle

Cleo moves prompts between Discovery, In Play and Earned as evidence changes. This keeps attention on questions where new work is most valuable. A stage change affects how often a prompt runs, not how it is scored. Retiring is always your choice, and a prompt never returns to Discovery.

​
Query fanouts

An AI engine may expand one user question into several searches or sub-questions. Fanout shows those related searches so you can understand what information the engine needed before composing its answer. Use fanouts to identify:
  • Missing supporting pages
  • Comparison and definition gaps
  • Evidence the answer expects from third-party sources
  • Variants of the question worth covering in one strong page
Fan-out coverage is the share of your prompts for which those searches were visible to Cleo. It is partial by nature, because only some engines publish the searches they ran.

​
Select AI engines

Cleo supports seven engines on SEO + GEO and SEO + GEO + Social plans: ChatGPT, Google AI, Bing AI, Perplexity, Gemini, Claude, and DeepSeek. Enterprise also makes Grok and Meta AI available, for nine available engines in total. A site can run up to four engines at the same time. Open GEO settings to review the active engines and slots for the site. Availability can also depend on rollout readiness and site configuration. If your plan permits changes, replace an engine deliberately: cross-engine scores are not interchangeable, and the new engine needs fresh runs before its trend becomes useful.