What is Large Language Model (LLM)?
The neural networks behind ChatGPT, Gemini, and Claude — trained on web-scale text, now mediating how buyers discover brands.
Definition
A large language model is a neural network trained on web-scale text to predict and generate language. GPT-5, Gemini, and Claude are LLMs; ChatGPT, AI Overviews, and Perplexity are products wrapping them. LLMs answer from two sources: parametric memory (patterns absorbed in training) and retrieval (live documents fed at answer time).
Why it matters
LLMs are becoming the interface between buyers and the web: AI platforms process 3.5B+ queries weekly, and half of B2B software buyers start vendor research in a chatbot (G2, 2026). For marketers the operational takeaway is the two-source model — what the model remembers about you (shaped over years by corroborated coverage) and what it retrieves about you (shaped this week by crawlable, answer-shaped pages). GEO works both.
Frequently asked
How do LLMs decide which brands to mention?
A blend of training-data association (which brands the corpus links to which problems) and retrieval-time evidence (which pages answer the prompt credibly). Corroboration across independent sources moves both.
Do different LLMs see my brand differently?
Often, yes — different training corpora, cutoffs, and retrieval stacks. That's why visibility tracking must run per-assistant rather than assuming one score.
Related terms
- llms.txtA proposed markdown file at /llms.txt that curates a site's key content for AI systems. Widely adopted, rarely fetched — know both facts.
- AI CitationsThe linked sources an AI assistant credits when generating an answer. Being cited is the AI-era equivalent of ranking #1.
- AI CrawlerBots operated by AI companies — GPTBot, ClaudeBot, PerplexityBot, Google-Extended — that fetch web content for model training or live answer retrieval.
- AI OverviewsGoogle's AI-generated summaries shown above traditional results, now appearing on roughly half of searches and dramatically reducing clicks to websites.