Jargon Buster

A deep learning framework that pits two neural networks against each other to generate new, synthetic data that closely resembles real data.

Generative AI (Gen AI) can create original content, for example text or images tailored to a request. It learns from data that it has been trained on, to produce new things. When it comes to digital advertising, Generative AI enables automated and optimised delivery of highly personalised content and experiences based on individual preferences, behaviours, and past interactions.

 The practice of improving how often a brand or content appears in AI-generated answers from tools like AI search or chat assistants. Instead of focusing on ranking in search results, it focuses on influencing what AI systems include in their responses.  

Often confused with Answer Engine Optimisation, in practice they overlap. AEO focuses on optimisation on the exact answer whereas GEO feeds the model that builds the answer. See more here.

The process of connecting an AI model to reliable, up-to-date information so its answers are based on real data rather than just what it learned during training. It helps make AI responses more accurate, relevant, and trustworthy. 
 
A retailer connects an AI assistant to its live product catalogue and pricing, so recommendations reflect current availability and offers. 

RAG is one technique used to achieve grounding, see more here.

Rules and safeguards that control how an AI system behaves to ensure its outputs are safe, accurate, and appropriate.  

A brand sets guardrails in an AI tool to ensure it only uses approved messaging, avoids risky claims, and stays consistent with brand tone.

Instances where a Generative AI model generates output that is not grounded in its input data, i.e., it "makes things up". This is particularly common in tasks like text generation from large language models, where the model might generate plausible sounding but incorrect or nonsensical information.

A system where people review, guide, or approve AI outputs before they are used or published. Instead of working fully on its own, the AI works alongside humans, combining automation with human judgement. 

An AI creates several ad variations, and a marketing team reviews and approves them before they go live to ensure quality and brand alignment. 

This helps balance efficiency with control, keeping AI outputs accurate, safe, and on-brand.

The process of an AI model using what it learned during training to generate predictions, decisions, or responses from new inputs. 
 
When you ask a chatbot a question and it responds, the model is performing inference by using what it learned during training to produce an answer in real time.

A large language model is a machine learning model that uses a huge amount of data to perform natural language processing tasks. LLMs can generate, translate, summarise, and predict text. They can also answer questions and perform other tasks. LLMs can automate tasks like content creation, personalise ads, and analyse customer data.