Glossary entry
Large Language Model Optimization (LLMO)
Also known as: LLMO, LLM Optimization
Optimisation of content and structured signals specifically for citation behaviour by large language models, irrespective of whether the LLM is wrapped in a search interface.
Large Language Model Optimization (LLMO) describes the practice of optimising content for the way LLMs treat sources — citation rate, attribution quality, brand mention rate — independent of whether the LLM is wrapped in a search interface.
The distinction from GEO is narrower than the literature suggests. In practice:
- GEO focuses on the search-engine surface (ChatGPT with browsing, Perplexity, AI Overviews) — the user asks a question, the engine searches the web, the engine cites sources in its answer.
- LLMO includes the same plus the no-browsing case: a user asks an LLM a question, the LLM answers from training data, and the question of "did the LLM remember and quote me correctly" is part of the surface.
Most practitioners treat LLMO and GEO as synonyms. We use GEO because the search-wrapped case is where the discipline produces measurable, repeatable signals; pure-training-data LLMO is harder to instrument.
See also: llms.txt: Specification, Adoption, and Practical Setup.