Jargon Buster

AI-generated audience groups that simulate real consumer behaviours and characteristics using aggregated data, without including personally identifiable information (PII). 
 
A media agency uses synthetic audiences to test campaign targeting strategies and predict performance without using real customer data, helping inform media planning decisions. 

Synthetic audiences enable advertisers to carry out research and planning in a privacy-safe way, supporting effective targeting and insight generation while reducing reliance on personal data. 

Content (text, images, video, audio) generated by AI rather than captured from real-world sources. It includes both entirely AI-generated content and hybrid content where AI enhances, edits or alters real-world media.  

A brand uses AI to generate personalised video ads featuring dynamic visuals and voiceovers tailored to different audience segments, without filming multiple versions. 
 
Synthetic media enables advertisers to create and adapt content at scale, making it easier to personalise campaigns, reduce production costs, and respond quickly to changing needs. 

A token is the basic unit of text that an AI model processes. A token can be a whole word, part of a word, or punctuation, depending on how the model breaks text into smaller pieces. 

AI models convert text into tokens so they can process it. The number of tokens a model can work with at one time is called its context window. See more here.

When a marketer asks an AI tool to write ad copy, both the prompt and the response are measured in tokens. The more tokens used, the more information the AI can process, which can also affect cost. 

Tokens determine how much information an AI model can process at once, influencing response length, performance, and, for many AI services, pricing. 

The dataset used to teach an AI model patterns, relationships, and behaviours during training. Typically consists of large volumes of structured or unstructured information such as text, images, audio, video, or tabular data, depending on the type of model being trained. 
 
An AI model trained on past campaign performance, audience data, and creative assets can learn what drives engagement and use those patterns to improve future campaigns.  
 
The quality and diversity of training data directly affect how accurate, relevant, and unbiased AI outputs are, making it a critical factor in advertising effectiveness.

A way of representing data (such as text, images, or audio) as numbers so an AI model can understand its meaning and compare it with other data. 
 
A recommendation engine uses vector embeddings to identify products or content that are similar in meaning, helping advertisers show more relevant ads to users based on their interests.  
 
Vector embeddings allow AI systems to understand relationships and similarity between content, improving search, targeting, recommendations, and personalisation in advertising. 
 

 A prompt-driven approach to building software where users describe what they want an AI system to create, then refine it through rapid iteration rather than traditional coding methods. 
 
A marketing team uses AI tools to quickly prototype a campaign landing page by describing the layout, messaging, and features, then iteratively refining the output without writing code from scratch.  
 
Vibe coding lowers the barrier to building digital experiences, allowing advertisers and non-technical teams to experiment, prototype, and launch ideas faster. 

The process of optimising ads and content for voice search, is an area where AI plays a significant role in understanding natural language queries.

AI models and techniques that provide human-understandable explanations of their decisions, enhancing transparency and trust.

A popular deep learning model for real-time object detection that can identify multiple objects in images or video.