Cluster 8 of 9

8. Model bias and inclusion dynamics

This is the most uncomfortable cluster in the hub. Why? Because it shows that GEO is not a pure content problem. Even with perfect on-site optimisation you can remain invisible in AI answers — in open category questions a small number of brands keeps returning, and whoever is not among them breaks in with difficulty. That pattern is what we call the inclusion threshold: a measurement concept, not a demonstrated model component.

This cluster explains how the inclusion threshold works, which biases sit in training data, and how you can systematically measure inclusion dynamics. It is also the place where GEO and brand strategy meet: the analogy with mental availability from brand science — broad and consistent presence, over time. An analogy, not a cognitive equivalent.

→ This cluster is strategic, not tactical. Start with 8.1 to understand the pattern before you go to measurement methods in 8.3.

In this section

8.1

The inclusion threshold: why some brands appear — and others don't

Some brands appear consistently in open category questions, others almost never — even with strong content. Inclusion threshold is the Groundbase name for that observable pattern. What it does and does not say, and how to work with it.

8.2

Training bias and frequency heuristics

Why do AI models reproduce market structure? Training bias, frequency heuristics and mental availability — the mechanisms behind AI visibility explained.

8.3

Measuring inclusion rate

How do you measure whether you reach the inclusion threshold? The open category prompt as method, the inclusion rate as metric, and benchmarking against competitors.