Foundation Models
Definition
Foundation models are large-scale AI models trained on broad, diverse datasets, giving them general capabilities across language, reasoning and sometimes images or audio, which can then be adapted, prompted or fine-tuned for many specific applications rather than being built for a single narrow task.
GPT, Claude, Gemini and Llama are examples of foundation models; the chat assistants and AI search products a brand is trying to be visible in are typically built on top of one or more of these models, combined with retrieval systems, tools and product-specific instructions.
Understanding foundation models helps explain why AI answers can vary between products even for the same question: each provider combines its foundation model with different retrieval sources, grounding rules and product design choices, which is why AI search visibility needs to be measured across multiple platforms, not just one.
BotRank's multi-LLM visibility analysis tracks a brand's presence across the products built on the major foundation models, since strong visibility on one does not guarantee the same on another.
Frequently Asked Questions
What is the difference between a foundation model and a chat assistant?
The foundation model is the underlying trained model; the chat assistant, such as ChatGPT or Claude.ai, is a product built on top of it, combined with retrieval, tools and interface design.
Why do different AI products answer the same question differently?
Because each product pairs its foundation model with different retrieval sources, grounding rules and product instructions, which affects what gets cited even when the underlying model is similar.
Are GPT, Claude and Gemini all foundation models?
Yes, each is a family of foundation models developed by a different company, each with its own training data, capabilities and update schedule.
Does a foundation model's training cutoff limit AI visibility?
It limits the model's static knowledge, but most AI search products pair the model with real-time retrieval, which is why up-to-date, well-structured web content still matters for visibility.
