Glossary/AI in Marketing

• Advantage+

  Meta's suite of AI-automated campaign tools for targeting, creative, and budget.

• Advantage+ Audience

  Meta's AI targeting that treats advertiser inputs as suggestions and finds the people most likely to convert beyond them.

• Advantage+ Creative

  Meta settings that automatically enhance and vary ad creative, such as adjusting images, text, and formats per viewer to improve performance.

• Advantage+ Sales Campaigns

  Meta's end-to-end automated sales campaign, renamed from Advantage+ Shopping in 2025, that uses AI to set audience, placement, budget, and creative.

• Agentic AI (AI Agents)

  AI systems that plan and take multi-step actions toward a goal with limited human input.

• Agentic Commerce

  Shopping in which AI agents research, compare, and buy on a consumer's behalf.

• Agentic Marketing

  Marketing in which AI agents carry out research, content, campaign execution, and optimization autonomously within goals and guardrails humans set.

• Agentic Workflow

  A multi-step process in which AI agents plan, use tools, check results, and iterate toward a goal rather than answering a single prompt.

• AI Acceptable Use Policy (AUP)

  A company policy defining which AI tools staff may use, for what tasks, and with what data, disclosure, and review rules.

• AI Avatar

  A realistic AI-generated or AI-animated human presenter used in videos and ads to deliver scripts without filming a real person.

• AI Bias

  Systematic unfairness in AI outputs, such as skewed targeting or stereotyped imagery, inherited from training data or model design.

• AI Bidding (Smart Bidding)

  Automated bid strategies that use machine learning to optimize toward a goal.

• AI Brand Voice

  Codified tone, vocabulary, and style rules fed to AI tools so generated content consistently sounds like the brand.

• AI Chatbot

  Software that converses with customers using natural language.

• AI Content Detection

  Tools that estimate whether text, images, or video were created by AI; useful but often unreliable, especially for edited text.

• AI Content Governance

  The policies, approval workflows, and checks that control how AI-generated marketing content is created, fact-checked, disclosed, and published on-brand.

• AI Copilot

  An AI assistant embedded in software that helps a person draft, analyze, or act inside that tool while the person stays in control.

• AI Dubbing

  Automatically translating and re-voicing video into other languages with AI, sometimes matching the original speaker's voice and lip movements.

• AI Evals

  Structured tests that measure how well an AI model or workflow performs on accuracy, brand fit, safety, and cost before and after launch.

• AI Governance

  Policies and controls for using AI responsibly, legally, and on-brand.

• AI Hallucination

  When an AI model states false or fabricated information as fact.

• AI Marketing Automation

  Using AI to run and optimize campaigns, workflows, and decisions with minimal manual work.

• AI Orchestration

  Coordinating multiple AI models, agents, tools, and data sources so they work together reliably in one business process.

• AI SDR

  An AI agent that does sales development work, such as researching prospects, writing outreach, and booking meetings, alongside or instead of human reps.

• AI Search (AI Overviews)

  Search experiences that return AI-generated summaries instead of, or above, links.

• AI UGC

  Ads made with AI to mimic user-generated content, typically an AI avatar talking to camera, used to test many hooks quickly and cheaply.

• AI Visibility (LLM Visibility)

  How often and how favorably a brand appears in AI assistant responses.

• AI Voice Cloning

  Using AI to replicate a specific person's voice from samples, for voiceovers, localization, or personalized audio, ideally with consent.

• AI Watermarking

  Embedding an invisible, machine-detectable signal in AI-generated content, such as Google's SynthID, to identify it as AI-made.

• AI Workflow Automation

  Connecting AI models to apps and data with tools like n8n, Zapier, or Make to run marketing tasks automatically end to end.

• AI-Generated Content

  Marketing copy, imagery, or video produced wholly or partly by AI.

• Amazon Ads Agent

  Amazon's agentic AI assistant, announced at unBoxed 2025, that recommends audiences and budgets, adjusts campaigns, and writes AMC queries from plain-language goals.

• Amazon Brand+ and Performance+

  AI-automated Amazon DSP campaign types using Amazon's predictive models: Brand+ for awareness and consideration, Performance+ for conversions.

• Amazon Creative Agent

  Amazon Ads agentic AI tool that researches a brand's products and generates display, video, audio, and streaming TV ad concepts and assets at no cost.

• Answer Engine Optimization (AEO)

  Structuring content so answer engines and assistants surface it directly.

• Autonomous Campaign Management

  Using AI to plan, launch, optimize, and reallocate ad budgets across campaigns with minimal human input, often via platform automation or AI agents.

• Brand Exclusions

  Google Ads control preventing Performance Max or AI-driven Search campaigns from serving on specified brand queries, such as your own or competitors' names.

• Chain-of-Thought

  Prompting or training an AI model to reason through intermediate steps before giving a final answer, improving accuracy on complex problems.

• Channel Performance Report

  Google Ads report, launched 2025, showing how Performance Max and Demand Gen results split across Search, YouTube, Display, Discover, Gmail, and Maps.

• Computer Vision

  AI that interprets images and video, used for visual search and contextual targeting.

• Content Authenticity (C2PA)

  Standards for labeling and verifying the origin of digital content, including AI-made assets.

• Context Engineering

  Designing everything an AI model sees, including instructions, examples, retrieved documents, memory, and tools, so it performs reliably; popularized in 2025.

• Context Window

  The maximum amount of text, measured in tokens, an AI model can consider at once, including the prompt, documents, and its own reply.

• Conversational Commerce

  Selling through chat interfaces, including messaging apps, brand chatbots, and AI assistants, where discovery and purchase happen in conversation.

• Conversational Marketing

  Engaging customers through chatbots and messaging in real time.

• Creative Automation

  Using AI and templates to produce, resize, localize, and version large volumes of ad creative from a few source assets.

• Custom GPT

  A tailored version of ChatGPT configured with specific instructions, files, and tools for a task, such as brand-voice copywriting.

• Deepfake

  AI-manipulated media that convincingly imitates a real person.

• Demand Gen Campaigns

  Google Ads AI-powered campaigns for visual ads across YouTube, Discover, and Gmail, aimed at creating demand and driving conversions.

• Diffusion Model

  The type of generative AI behind most image and video generators, which builds visuals by gradually refining random noise.

• Embeddings

  Numerical representations of text, images, or products that capture meaning, letting software find similar items even when wording differs.

• Few-Shot Prompting

  Including a handful of examples in a prompt so an AI model copies the desired format, tone, or reasoning.

• Final URL Expansion

  Google Ads automation that swaps your chosen landing page for a more relevant page on your site and generates matching ad text.

• Fine-Tuning

  Further training a model on specific data to adapt it to a task or brand voice.

• Foundation Model

  A large AI model trained on broad data, such as GPT, Gemini, Claude, or Llama, that can be adapted to many tasks.

• Generative Ads

  Ads whose images, video, copy, or variations are created by generative AI, often automatically inside platforms like Meta, Google, or TikTok.

• Generative AI

  AI that creates new text, images, audio, or video from prompts.

• Generative Engine Optimization (GEO)

  Optimizing content so it is cited and recommended in AI-generated answers.

• Generative Fill

  An AI editing feature that adds, removes, or extends parts of an image from a text prompt, popularized by Adobe Photoshop.

• Guardrails

  Rules and filters that keep AI outputs safe, accurate, on-brand, and compliant, such as blocked topics, claim checks, and approval steps.

• Human-in-the-Loop (HITL)

  A workflow where people review, approve, or correct AI outputs or agent actions before they take effect.

• Hyper-Personalization

  Using AI and real-time data to tailor content to an individual.

• Inference

  The act of running a trained AI model to produce an output; inference costs drive the price of AI features at scale.

• Jailbreak

  A prompt crafted to get an AI model to bypass its safety rules and produce content it is designed to refuse.

• Knowledge Cutoff

  The date after which an AI model has no training data, so it cannot know later events unless connected to search or other live sources.

• Knowledge Graph

  A network of entities and their relationships, like brands, products, and people, that helps search engines and AI systems understand facts.

• Large Language Model (LLM)

  An AI model trained on vast text to understand and generate language.

• LinkedIn Accelerate

  LinkedIn's AI campaign builder that generates targeting and ad copy from a URL and automatically optimizes B2B campaigns.

• llms.txt

  A proposed file that tells AI models which site content is most useful to read.

• Low-Rank Adaptation (LoRA)

  An efficient fine-tuning technique that trains small add-on weight layers instead of the whole model, cheaply adapting models to a brand's style or products.

• Machine Learning

  Algorithms that learn patterns from data to make predictions or decisions.

• Meta Andromeda

  Meta's AI ad-retrieval engine, introduced December 2024, that narrows millions of candidate ads to a shortlist per person, rewarding diverse creative.

• Meta GEM (Generative Ads Recommendation Model)

  Meta's LLM-scale foundation model for ads that learns patterns across its apps and transfers that knowledge to the models ranking ads in real time.

• Meta Lattice

  Meta's ad-model architecture that consolidates many smaller ranking models into fewer large shared ones so learnings transfer across objectives.

• Model Card

  A standardized document describing an AI model's intended uses, training data, performance, and limitations, used to evaluate vendors and risks.

• Model Collapse

  Degradation that occurs when AI models are trained heavily on AI-generated content, causing outputs to grow less diverse and accurate.

• Model Distillation

  Training a smaller 'student' model to mimic a larger 'teacher' model, delivering similar quality at lower cost and latency.

• Multi-Agent System

  A setup where several specialized AI agents, such as research, copy, and analytics agents, coordinate to complete a larger task.

• Multimodal AI

  AI that can understand and generate more than one type of input or output, such as text, images, audio, and video.

• Natural Language Processing (NLP)

  The field of AI that lets computers understand and generate human language.

• Next-Best Action

  An AI recommendation of the most valuable next interaction for each customer.

• On-Device AI

  AI models that run locally on a phone, laptop, or other device rather than in the cloud, improving privacy and speed, as in Apple Intelligence.

• Open-Weight Model

  An AI model whose trained parameters are publicly released, such as Llama or Mistral, so companies can host and customize it themselves.

• Performance Max (PMax)

  Google's AI-driven campaign type that runs across all Google channels from one campaign.

• Pinterest Performance+

  Pinterest's AI and automation ad suite, launched 2024, that automates targeting, bidding, budgets, and creative with about half the setup inputs.

• Predictive Analytics

  Using data and models to forecast outcomes such as churn, purchase, or LTV.

• Prompt Engineering

  Crafting instructions to get reliable, high-quality output from AI models.

• Prompt Injection

  An attack where hidden or malicious text tricks an AI model or agent into ignoring its instructions or leaking data.

• Propensity Modeling

  Scoring how likely each customer is to take an action, such as buying or churning.

• Quantization

  Compressing an AI model by storing its weights at lower numerical precision, cutting memory and cost so it can run faster or on devices.

• Reasoning Model

  An AI model that works through problems step by step before answering, trading speed and cost for better accuracy on complex tasks.

• Reinforcement Learning from Human Feedback (RLHF)

  Training method where humans rank model outputs and a reward model uses those rankings to tune an LLM toward helpful, safe responses.

• Responsible AI

  Principles and practices for building and using AI fairly, transparently, safely, and accountably, including bias checks and human oversight.

• Retrieval-Augmented Generation (RAG)

  Grounding an AI model's answers in retrieved documents or data.

• Search Themes

  Performance Max signals, up to 50 per asset group, telling Google AI which search topics matter to your business, similar to keywords.

• Semantic Search

  Search that matches the meaning and intent of a query rather than exact keywords, typically using embeddings.

• Sentiment Analysis

  Using AI to classify text as positive, negative, or neutral toward a brand or topic.

• Shadow AI

  Employees using unapproved AI tools at work, creating risks of leaking confidential or customer data and off-brand outputs.

• Share of Model

  A brand's share of mentions or recommendations across AI model answers in its category.

• Small Language Model (SLM)

  A compact language model that runs cheaply and quickly, sometimes on-device, for focused tasks where a large model is unnecessary.

• Synthetic Audiences (Synthetic Personas)

  AI-simulated customer panels used to test messages and concepts.

• Synthetic Data

  Artificially generated data that mimics real data for modeling without exposing individuals.

• Synthetic Media

  Any image, video, audio, or text generated or substantially altered by AI, from harmless product shots to deceptive deepfakes.

• System Prompt

  Standing instructions given to an AI model that set its role, tone, rules, and limits before any user message.

• Temperature

  A model setting that controls randomness: low values give consistent, predictable output, high values give more varied, creative output.

• Text-to-Image

  Generative AI that creates images from a written prompt, using tools like Midjourney, DALL-E, Firefly, or Imagen.

• Text-to-Video

  Generative AI that creates video clips from written prompts or images, using models such as Sora, Veo, Runway, or Kling.

• TikTok Smart+

  TikTok's AI performance campaign suite, launched October 2024, that automates targeting, bidding, and creative, with full or partial automation options.

• TikTok Symphony

  TikTok's generative AI creative suite for advertisers, offering script writing, AI avatars, video generation, and automatic creative enhancements.

• Token

  The small chunk of text, often part of a word, that AI models read and generate; usage, limits, and pricing are measured in tokens.

• Tool Calling

  An AI model's ability to trigger external functions, such as searching, querying a CRM, or sending an email, instead of only writing text.

• Vector Database

  A database that stores embeddings and quickly finds the most similar ones, powering semantic search, recommendations, and RAG.

• Virtual Influencer

  A computer-generated social media persona that brands partner with or create to promote products like a human influencer.

• Zero-Shot Prompting

  Asking an AI model to do a task with instructions only and no examples, relying on what it already knows.