Separate the experiences

Google describes AI Overviews as summaries in Search and AI Mode as an experience for deeper exploration and complex questions. They may use different models and techniques, so their answers and links can differ. Source: Google Search Central.

Compare like with like

Create separate records for the same question in each experience. Keep the question wording, collection time and any known language or location context attached. Do not merge the answers into a single Google visibility number before deciding what that number means.

In PromptFace, the public label Google AI Model refers to Google AI Mode. Google AI Overview is a separate identity. Use the identity on the record when reading or exporting observations.

Record absence accurately

An Overview that does not appear is a different observation from an Overview that appears without your brand. A failed collection is different again. Keep these states separate so a service error does not look like lost visibility.

For example, define one field for whether an answer was available and another for whether the brand appeared. Only calculate brand presence among the observations that meet your stated inclusion rule.

Turn differences into questions

If a source appears in one experience only, inspect what it contributes: a product fact, a comparison or a direct answer. Write down a hypothesis you can investigate. Do not infer a universal ranking rule from a single pair of answers.

Keep the original response and source context beside your analysis. That makes a later review possible when the live search result or the source page changes.