LLMO

LLMO stands for Large Language Model Optimization and refers to measures with which companies work towards appearing in the outputs of large language models. The term is often used synonymously with GEO; where a distinction is made, LLMO puts more emphasis on the models themselves and GEO on the search interfaces built on top of them.

Two paths should be distinguished, which are readily mixed up in everyday use. One concerns answers where the system looks things up on the web at runtime — the same levers apply here as for GEO. The other concerns the knowledge stored in the model from training, which can practically not be influenced in the short term and which is created anew with every new model version.

Serious work on LLMO therefore limits itself to the first path and to building verifiable mentions in sources that qualify for training data. Offers that promise a fixed placement in model knowledge greatly overestimate how controllable it is.