LLM (Large Language Model)
A neural model trained on vast text that predicts language — the engine behind every AI answer.
What is an LLM?
A Large Language Model (LLM) is a neural network trained on vast amounts of text to predict and generate language. LLMs power the AI engines that answer queries — ChatGPT, Claude, Gemini and Perplexity.
The LLM is the substrate of AEO/GEO: if you want a brand cited inside AI answers, you are optimising for how an LLM selects and phrases sources.
An LLM learns statistical patterns of language from a training corpus, encoding that knowledge in billions of parameters (its weights). At inference it generates text token by token, each choice conditioned on the prompt and everything generated so far. This is why the same question can yield slightly different answers across runs — a property SkuLift handles with N-sampling.
For visibility, two LLM behaviours matter. First, parametric recall: an answer drawn purely from the weights, with no live lookup. Second, retrieval-augmented generation, where the model is fed fresh sources before answering. A brand can be cited in either mode, and SkuLift measures both separately because the levers differ.
Because every answer engine is built on an LLM, LLM visibility — being mentioned and cited in generated answers — is the new surface to compete on. Classic SEO ranked you in a list of links; LLM-era optimisation gets you named inside the answer itself.