When you run a Gumshoe report, different AI models behave in different ways. Some use live web search to fetch information in real time, while others rely primarily on their foundational training data. Both are important for understanding your visibility.
Live Search Models
These models actively query the web at the time of the report run.
Why they matter:
- Capture fresh, real-world web content (including your newest pages and press mentions).
- Show how AI systems are citing live domains and categories.
- Help you identify the sources driving the influence right now.
Foundational Training Models
These models rely mostly on their built-in training data and reasoning.
Why they matter:
- Show how AI recommends brands by default, even without checking the live web.
- Reflect the underlying narratives and biases baked into each model’s training.
- Provide visibility into how your brand will appear in offline or default LLM responses, which still power many AI assistants.
Why You Need Both
- Using only live search models can show you recency, but misses how users are experiencing AI Search.
- Using only foundational models shows default reasoning but ignores how fresh citations can shift outcomes.
- AI models increasingly initiate live search only when a user explicitly selects it or when it’s required for accuracy, which means many of their responses still rely on their foundational training data rather than live search results.
By running reports across both types, Gumshoe gives you:
- A balanced picture of static brand perception (training) vs. dynamic influence (search).
- Clear insight into how to optimize both your long-term reputation and your real-time visibility.
- The ability to be prepared when the AI Model foundational training is updated.
How do I know which type of model I'm selecting?
When selecting models in Gumshoe, tooltips indicate whether each model uses live search or foundational knowledge, so you know what you're measuring.
⚡ Pro Tip: Foundational models are slower to change, while live search models can shift daily. Scheduled multi-model reports help you track both patterns effectively.