Glossary · Beginner
What Is a LLM? Definition and Example
An LLM (large language model) is an AI model trained on very large amounts of text so it can predict the next words in a sequence, and through that, answer questions, summarise and write. Products such as ChatGPT, Gemini, Claude and Perplexity are built on LLMs. For marketers, the LLM is the part that writes the answer your buyer reads.
How does an LLM work, in plain terms?
During training, the model sees huge volumes of text and learns statistical patterns: which words, facts and ideas tend to go together. When you ask a question, it generates a response one piece of text at a time, based on those patterns and on whatever extra context it's been given.
That extra context matters. Many AI products now retrieve web pages at the moment you ask and give them to the model, an approach researchers call retrieval-augmented generation (Lewis et al., arXiv). That's why answers can include current information and source links that weren't in the training data.
LLM vs AI product: what's the difference?
| LLM | AI product (e.g. ChatGPT) | |
|---|---|---|
| What it is | The trained model | An application built around one or more models |
| Knows recent events? | Only up to its training data | Can, if it uses web search |
| Shows sources? | No | Often, when it retrieves pages |
| What you can influence | Indirectly, via public content over time | Retrieval: crawler access, indexable pages, cited sources |
Example
Illustrative. A buyer asks an assistant "what's a good invoicing tool for freelancers?". Without web search, the answer comes from patterns the model learned in training, which may be out of date. With search on, the product fetches review sites and vendor pages, and the LLM writes an answer that blends both and cites some of the pages.
Why does it matter for GEO?
You can't edit what a model learned, but you can make sure the pages it retrieves are accessible, accurate and easy to quote. That means allowing the right search crawlers (for example OAI-SearchBot for ChatGPT search, per OpenAI), keeping key facts in plain HTML text, and earning mentions on the sources engines already cite.
Common misconceptions
- "The LLM looked me up." Only if the product used retrieval; otherwise it's recalling patterns.
- "Training on my site guarantees mentions." No mechanism guarantees how a model uses any page.
- "All AI tools use the same model." They differ, and so do their answers.
Related
Questions
What does LLM stand for?
Large language model, a model trained on large amounts of text to understand and generate language.
Are ChatGPT and Gemini LLMs?
They are products built on LLMs, combined with other systems such as web search, tools and safety layers.
Do LLMs search the web?
The model itself doesn't. Products around it can retrieve web pages and pass them to the model, which is how cited answers are produced.