AI Glossary

The AI marketing glossary, in plain language.

Plain-language definitions of 138 AI terms and tools marketing teams run into most, from agentic AI to zero-shot prompting.

AI Models & Chatbots 7 terms

ChatGPT
OpenAI's conversational AI assistant powered by the GPT family of large language models. Used for drafting, research, summarization, and coding. The most widely adopted AI tool in marketing as of 2024–2025.
Claude
Anthropic's family of large language models known for long context windows, careful reasoning, and strong writing. Frequently used for editorial work, research, and AI-assisted content systems.
Gemini
Google's family of multimodal AI models, integrated into Google Search (AI Overviews), Workspace, and the Gemini chatbot. Known for native multimodality and fast factual answers.
Copilot
Microsoft's AI assistant, embedded across Windows, Office, and Edge. Built on OpenAI models for most consumer use, with enterprise-grade Microsoft 365 Copilot for business workflows.
Llama
Meta's open-weight large language models. Frequently fine-tuned by enterprises that want to host AI in-house rather than rely on a closed API.
Grok
xAI's conversational AI, integrated with X (formerly Twitter). Pulls real-time information from the platform.
Mistral
Open-source AI models from a French AI lab. Popular for self-hosted enterprise deployments and developer-grade AI products.

AI Search & Answer Engines 6 terms

Perplexity
An AI answer engine that combines a chatbot with cited web search results. Returns a direct answer plus the sources it pulled from. The reference implementation of AEO.
Google AI Overviews
AI-generated answer summaries shown at the top of Google search results pages, replacing traditional featured snippets for many queries. Formerly called Search Generative Experience (SGE).
You.com
An AI-first search engine that returns a chatbot-style answer alongside traditional results.
Phind
A developer-focused AI search engine built around code and technical documentation.

Marketing AI Tools 9 terms

Jasper
A marketing-focused AI writing platform with brand voice and template features built for content teams.
Copy.ai
AI writing platform aimed at marketers and sales teams, with workflow automation for outbound and content production.
Writer
Enterprise AI platform with custom-trained models, brand-voice enforcement, and governance features for in-house marketing teams.
Notion AI
AI features built directly into Notion documents — drafting, summarization, translation, and Q&A across a workspace.
HubSpot AI
AI features inside the HubSpot marketing, sales, and service platform — content drafting, predictive analytics, and chatbot building.
Surfer SEO
Content optimization tool that uses AI to analyze top-ranking pages and recommend on-page improvements.
Frase
AI-assisted SEO and content brief tool that turns search queries into structured outlines and optimized drafts.
Zapier
Workflow automation platform with AI-powered actions, AI agents, and chatbot integrations across thousands of apps.
Smash Balloon
A WordPress plugin family for embedding live social feeds (Instagram, Facebook, Twitter, YouTube) into a website.

AI Image & Video Generation 6 terms

DALL·E
OpenAI's image generation model, integrated into ChatGPT for prompt-based image creation.
Midjourney
A high-quality generative image tool known for stylized, editorial output. Widely used by brand designers.
Stable Diffusion
Open-source image generation model that runs locally or on cloud infrastructure. Highly customizable.
Adobe Firefly
Adobe's family of generative AI models, integrated into Photoshop, Illustrator, and Express. Trained on commercially licensed content.
Sora
OpenAI's text-to-video generation model, designed for short-form cinematic clips from a written prompt.
Runway
AI video generation and editing platform popular with filmmakers and creative directors.

Key AI Terms 100 terms

Agentic AI
AI that pursues a goal over several steps on its own: planning, using tools and checking its work, rather than answering one prompt at a time.
AI agent
An AI system that takes multi-step actions for a user, often combining browsing, tools and memory.
AI assistant
A chat-based AI that helps with everyday tasks like drafting, research and summarizing. ChatGPT, Claude and Gemini are the best-known examples.
AI bias
Skewed or unfair output caused by gaps or imbalances in a model's training data. A real brand risk in audience targeting, imagery and copy.
AI disclosure
Telling your audience when content was created or substantially assisted by AI. Some platforms and regulations now require it.
AI governance
The policies, roles and review processes that decide how an organization uses AI, who approves what, and how risk is managed.
AI guardrails
Policies, prompts and review processes that keep AI output brand-safe, on-voice and factual.
AI localization
Using AI to adapt content for different languages, regions and cultures, beyond word-for-word translation.
AI personalization
Using AI to tailor messages, offers or content to individual people or segments at scale.
AI policy
A written acceptable-use policy that tells employees which AI tools they can use, for what, and with which data.
AI slop
Generic, low-effort AI content that sounds like everything else online. The reason brand voice guardrails matter.
AI watermarking
Invisible signals embedded in AI-generated images, audio or text so they can later be identified as AI-made.
API
Application Programming Interface: the connection that lets software send requests to an AI model directly, without a chat window. How AI gets built into other tools.
Artificial General Intelligence (AGI)
A hypothetical AI that can match humans across nearly any intellectual task. Today's tools are not AGI.
Artificial intelligence (AI)
Computer systems that perform tasks that normally require human intelligence, such as understanding language, recognizing images and making predictions.
Benchmark
A standardized test used to compare AI models on skills like reasoning, coding or writing. Useful, but not a substitute for testing on your own work.
Chain-of-thought
Prompting a model to work through a problem step by step before answering, which often improves accuracy on complex tasks.
Chatbot
Software that holds a text or voice conversation with people. Modern chatbots are powered by large language models.
Chunking
Splitting long documents into smaller pieces so an AI system can search and retrieve the most relevant parts.
Classification
Using AI to sort content into categories, such as tagging support tickets, comments or leads by topic or intent.
Closed model
An AI model whose inner workings and weights are private and only available through the provider's app or API, like GPT or Claude.
Computer vision
The field of AI that interprets images and video, powering features like image tagging, product recognition and visual search.
Content provenance (C2PA)
An industry standard that attaches tamper-evident Content Credentials to media, showing how it was made and whether AI was involved.
Content repurposing
Using AI to turn one anchor piece, such as an article or webinar, into many formats: posts, emails, scripts and summaries.
Context engineering
Deciding what information, examples and instructions to give an AI model so it has exactly what it needs to do a task well.
Context window
How much text a model can consider at once, measured in tokens.
Conversational AI
AI designed to understand and respond in natural back-and-forth dialogue, by text or voice.
Copilot
An AI assistant built into software you already use, working alongside you rather than on its own. Also the name of Microsoft's assistant.
Custom GPT
A no-code, customized version of ChatGPT set up with your own instructions, files and voice for a specific job.
Data privacy
Protecting personal and confidential information when using AI tools, including what gets pasted into prompts and how vendors store it.
Deep learning
A type of machine learning that uses many-layered neural networks. It is the foundation of modern generative AI.
Deepfake
Realistic AI-generated video, audio or images that impersonate a real person. A growing reputational and security risk for brands.
Diffusion model
The type of AI behind most image generators. It learns to build a picture by gradually removing noise from random static.
Embedding
A numeric representation of text that captures meaning. Used to power semantic search, recommendations and RAG systems.
Evals
Tests you run on AI output to check quality, accuracy and brand fit before and after rolling out a tool or prompt.
Explainability
How well people can understand why an AI system produced a particular output or decision.
Few-shot prompting
Including a few examples in your prompt so the model copies their style or format. One of the easiest ways to get on-brand output.
Fine-tuning
Further training a base model on a custom dataset so it adopts a specific tone, domain or task.
Foundation model
A large, general-purpose AI model trained on broad data that can be adapted to many tasks. GPT, Claude, Gemini and Llama are foundation models.
Function calling
A model's ability to trigger outside tools, like a calendar, CRM or search engine, by producing a structured request. Also called tool use.
Generative AI
AI that produces new content (text, images, audio, video, code) rather than only classifying existing data.
GPT
Generative Pre-trained Transformer: OpenAI's family of language models, and the technology behind ChatGPT.
Grounding
Connecting AI answers to trusted sources, such as your own documents or live search, so responses are factual and specific.
Hallucination
When AI produces information that sounds plausible but is wrong or made up. A core risk in unreviewed AI content.
Human-in-the-loop
A workflow where a person reviews or approves AI output before it's used or published.
Inference
The moment a trained model generates an answer to a prompt. Inference speed and cost affect how practical a tool is at scale.
Inpainting
Editing part of an image with AI, such as removing an object or changing a background, while keeping the rest intact.
Jailbreak
A prompt designed to trick an AI model into ignoring its safety rules. A risk for public-facing brand chatbots.
Knowledge base
A collection of approved documents, FAQs and guidelines that an AI system can search to answer questions accurately.
Knowledge cutoff
The date after which a model has no training data. Without web search, it won't know about anything more recent.
Knowledge graph
A map of entities (people, brands, products) and how they relate. Search and AI engines use knowledge graphs to understand who you are.
Large Language Model (LLM)
An AI system trained on massive text datasets to predict and generate human-like text. The technology behind ChatGPT, Claude and Gemini.
Latency
How long an AI system takes to respond. It matters for live chat, voice and real-time personalization.
Machine learning
A branch of AI where systems learn patterns from data instead of following hand-written rules.
Memory
A feature that lets an AI assistant remember facts and preferences across conversations.
Model Context Protocol (MCP)
An open standard, introduced by Anthropic, for connecting AI assistants to outside tools and data sources in a consistent way.
Multi-agent system
Several AI agents working together, each with its own role, such as a researcher, a writer and an editor.
Multimodal AI
AI that handles several input and output types (text, image, audio, video) in a single model.
Named entity recognition
AI that picks out names of people, companies, places and products in text. Useful for monitoring mentions and structuring content.
Natural language processing (NLP)
The field of AI focused on understanding and generating human language.
Neural network
A computing system loosely modeled on the brain, made of layers of connected nodes that learn patterns from data.
Open-weight model
A model whose trained weights are publicly released so companies can run and customize it themselves, like Llama or Mistral.
Orchestration
Coordinating multiple AI models, tools and steps into one reliable workflow.
Parameters
The internal values a model learns during training. Model size is often described in parameters, from millions to trillions.
Personally identifiable information (PII)
Data that can identify a person, like names, emails or account numbers. It should never go into AI tools without approved safeguards.
Predictive analytics
Using AI and historical data to forecast outcomes, such as which leads will convert or which customers may churn.
Prompt
The instruction or question you give an AI model. Clear context, audience and examples make prompts dramatically better.
Prompt chaining
Breaking a big task into a sequence of prompts, where each output feeds the next step.
Prompt engineering
Writing instructions to a generative AI model to get the output you want. A core skill for marketers.
Prompt injection
Hidden instructions planted in content, like a web page or document, that try to hijack what an AI system does.
Prompt library
A shared, organized collection of tested prompts your team reuses so output stays consistent and on-brand.
Prompt template
A reusable prompt with blanks to fill in, such as audience, channel and topic, so anyone on the team gets reliable results.
Reasoning model
An AI model built to think through problems step by step before answering. Stronger on analysis and planning, usually slower.
Recommendation engine
AI that suggests products, content or next actions based on behavior and preferences, like Netflix or Amazon recommendations.
Red teaming
Deliberately trying to make an AI system fail or misbehave to find risks before customers do.
Reinforcement learning from human feedback (RLHF)
A training method where people rate model answers and the model learns to prefer the better ones. It's part of why chatbots feel helpful.
Responsible AI
Designing and using AI in ways that are fair, transparent, safe and accountable.
Retrieval
Finding and pulling the most relevant information from a knowledge base so an AI model can use it in its answer.
Retrieval Augmented Generation (RAG)
The model pulls relevant documents from a knowledge base before answering, which reduces hallucination.
Sentiment analysis
AI that detects whether text is positive, negative or neutral. Used for social listening and review monitoring.
Small language model (SLM)
A compact language model that is faster and cheaper to run, often good enough for focused tasks.
Speech-to-text
AI that turns spoken audio into written text. It powers transcription, captions and meeting notes.
Summarization
Using AI to condense long content, like reports, calls or articles, into key points.
Synthetic data
Artificially generated data used to train or test AI when real data is scarce or sensitive.
Synthetic media
Images, video, audio or text created by AI rather than captured or written by people.
System prompt
Standing instructions given to an AI model before a conversation starts, setting its role, rules and voice.
Temperature
A setting that controls how predictable or creative a model's output is. Lower is more consistent; higher is more varied.
Text-to-image
AI that creates images from a written description. Midjourney, DALL·E and Adobe Firefly are examples.
Text-to-speech
AI that turns written text into natural-sounding spoken audio.
Text-to-video
AI that generates video clips from a written description, like Sora or Runway.
Token
The unit of text a model reads and writes. Roughly 3–4 characters, or 0.75 words in English.
Top-p
A setting that limits a model's word choices to the most likely options. Like temperature, it trades creativity for consistency.
Training data
The text, images or other content a model learns from. Its quality and coverage shape everything the model produces.
Transformer
The neural network design behind modern language models. The T in GPT.
Upscaling
Using AI to increase an image's or video's resolution while adding realistic detail.
Vector database
A database built to store and search embeddings. The infrastructure behind most AI knowledge bases.
Voice cloning
AI that reproduces a specific person's voice from a sample recording. Always get consent and disclose its use.
Zero-shot prompting
Asking a model to do a task with no examples, relying only on your instructions.

AEO & Search Terms 10 terms

Answer Engine Optimization (AEO)
The discipline of structuring brand content, schema, and topic clusters so AI answer engines (ChatGPT, Perplexity, Google AI Overviews, Claude) cite the brand in their responses. AEO is replacing classic SEO as the primary discovery channel for considered purchases.
Generative Engine Optimization (GEO)
Synonym for AEO, more common in academic and SEO-industry literature.
Schema markup
Structured data added to a web page in JSON-LD format, telling search engines and AI engines what the page is about (Person, Organization, Article, FAQPage, etc.).
Entity SEO
An approach to SEO that optimizes for clearly defined entities — people, places, brands, products — rather than only keywords. Foundational to AEO.
Topical authority
The cumulative signal that a brand or domain is an expert on a topic, built through depth of coverage, internal linking, and third-party citations.
Knowledge panel
The summary card Google shows on the right side of search results for a recognized entity. A signal of strong entity SEO.
E-E-A-T
Google's evaluator framework: Experience, Expertise, Authoritativeness, Trustworthiness. AI engines weigh similar signals when choosing brands to cite.
Brand voice
The consistent tone, vocabulary, and personality a brand uses across its content. The hardest thing for AI to fake — and the thing AEO depends on.

How does Marji use this glossary in client work?

Every AI-Augmented Social Strategy and AI Workshops & Advisory engagement starts with a shared vocabulary. This glossary becomes a living document inside client engagements, extended for the brand's own tools and reviewed quarterly.

Missing a term? Suggest an addition →

Ready to make your marketing AI-forward?

Marji takes on a small number of partner engagements each quarter.

Request a complimentary discovery call → Send a brief