This builds on Mike King’s excellent patent analysis (ipullrank.com). The next step: making the inner workings of Google’s AI Mode visible in real-world usage.
Using a targeted method – internally referred to at GPT Insights as the Core Tip – we’ve been able to prompt Google’s AI Mode to reveal its decision logic:
- which search queries are generated during fan-out
- how sources are evaluated
- which results are prioritized
- how search intent is interpreted
- which user context (location, language, device, situation) is assumed
I’ve applied this method to a broad range of prompts. The result:
A consistent yet variably detailed picture of how Google’s AI Search operates – depending on prompt and intent.
Sometimes the system explains how queries are formed, sometimes why specific sources are selected. Over time, a fuller understanding emerges.
The real value:
What was once only inferred from patents now shows up in live system behavior – concrete, traceable, and strategically usable.
More to come soon at GPT Insights.
