Insights · Glossary
AI and marketing glossary
79 common AI and marketing terms, each with a one-sentence definition, an explanation and an example. Open a term for the full entry.
AI Terminology
LLM Basics
What large language models are, how to talk to them, and where they make mistakes.
- Large Language Model (LLM)
- AI models trained on massive volumes of text that can understand and generate text. Doubao, DeepSeek, ChatGPT, and Claude are all powered by large language models.
Full entry → - Generative AI (AIGC)
- AI capable of generating new content such as text, images, audio, and video, as well as the content produced by such AI.
Full entry → - Prompt
- The instructions and reference materials provided to an AI, specifying what it should do, what information to use, and what format to follow.
Full entry → - System Prompt
- Pre-set instructions configured before a conversation begins that define the AI's role, rules, and response style. Users typically do not see them.
Full entry → - Context window
- The maximum total amount of text an LLM can process at once, including prompts, reference materials, chat history, and its own generated responses.
Full entry → - Token
- The basic unit LLMs use to process text. Both model context length and API call costs are calculated in tokens.
Full entry → - Hallucination
- When an AI presents non-existent or incorrect information with full confidence, such as fabricating data, sources, legal clauses, or product features.
Full entry → - Temperature
- A parameter that controls the randomness of an LLM's responses. Lower values produce more stable answers, while higher values lead to more varied outputs.
Full entry → - Multimodal
- AI can simultaneously process multiple types of information, including text, images, audio, and video.
Full entry → - Reasoning Model
- A large language model that internally analyzes a problem step by step before delivering a conclusion, making it ideal for calculations, planning, and complex judgments.
Full entry → - Pre-training
- The stage where a large language model learns language and general knowledge from massive volumes of public text, determining its baseline capabilities and knowledge cutoff date.
Full entry → - Fine-tuning
- Training an already pre-trained foundation model further on a company's own dataset so it adapts to a specific task or writing style.
Full entry → - Vector (Embedding)
- Converting a block of text or an image into a series of numbers. Content with similar meanings gets similar numbers, making it possible to retrieve information by meaning.
Full entry → - Text-to-Image
- The AI capability to generate images from a text description, or generate new images based on a reference image (image-to-image).
Full entry → - API
- Interfaces that allow software programs to interact with each other. Enterprise systems and AI agents use APIs to access capabilities such as large language models, search, image generation, and publishing.
Full entry → - Vibe Coding
- Instead of writing code yourself, you chat with an AI to have it write code, guiding it step-by-step through development and deployment.
Full entry →
AI Glossary
AI Agents and Knowledge Bases
Get AI to actually finish the job: how AI agents, skills, tools, and company data work together.
- AI Agent
- An AI that can read files, call external tools, and complete multi-step tasks on its own, unlike a standard chat AI that only replies with text.
Full entry → - The Three Pillars of AI Adoption
- Starbox's framework for putting AI to work: AI Agents do the work, SKILLs standardize how each task is handled, and Knowledge Bases store your company's data and institutional experience.
Full entry → - SKILL
- A standardized operating procedure written for an AI agent that defines how to run a specific task, the sequence of steps, which materials to use, and the final delivery format.
Full entry → - MCP
- An open protocol that connects AI to external tools and data, enabling AI agents to read and write to databases and trigger business systems.
Full entry → - Tool Calling (Function Calling)
- As an LLM generates a response, it calls external tools as needed—such as search, calculations, database queries, or messaging—and continues working based on the results.
Full entry → - Workflow
- Break a task into fixed steps and assign them in sequence to AI or humans, with predefined inputs and outputs for each step.
Full entry → - Multi-Agent Collaboration
- Divide a complex task among multiple specialized AI agents, coordinated by an orchestrator agent that assigns tasks and compiles results.
Full entry → - Human-in-the-Loop
- Keep human approval checkpoints within AI-driven workflows: humans make critical decisions, while AI handles preparation and execution.
Full entry → - Knowledge Base
- A repository of company documents and institutional knowledge that AI can search and reference, enabling it to answer questions and produce content grounded in internal facts.
Full entry → - RAG (Retrieval-Augmented Generation)
- A method where AI retrieves relevant information from a knowledge base first, then generates an answer based on those materials.
Full entry → - Vector Database
- A database that stores vectors and quickly finds semantically similar content, commonly used as a core component for RAG and knowledge base retrieval.
Full entry → - Text Chunking
- Splitting long documents into appropriately sized chunks before saving them to a knowledge base, making it easy to retrieve only the relevant sections.
Full entry → - LLM Wiki
- One-page referenced briefs compiled by AI from notes, files, and meeting transcripts to answer 'What has been decided?' and 'Which version is the latest?'
Full entry → - Long-term memory
- Information an AI retains across multiple conversations, such as user preferences, past decisions, and completed actions.
Full entry → - On-premises deployment
- Installing AI systems or models on an enterprise's own servers or dedicated cloud environment so data never passes through public platforms.
Full entry → - AI governance
- An enterprise's rules and oversight for AI usage: who can use it, what data it can handle, who is accountable for errors, and how actions are logged and audited.
Full entry →
AI Terms
AI Search and GEO
Users now ask AI directly. How brands appear in AI answers.
- GEO (Generative Engine Optimization)
- The optimization method to get brands and content retrieved, trusted, and included in answers by AI engines, covering generative AI platforms such as Doubao, DeepSeek, ChatGPT, and Perplexity.
Full entry → - SEO (Search Engine Optimization)
- The optimization method to rank webpages higher in search engine result lists, measured by rankings and clicks.
Full entry → - AI Search
- Users ask AI directly, and the AI synthesizes an answer from multiple sources instead of returning a list of webpage links.
Full entry → - Web Search
- A feature where AI searches the web for the latest content before answering, then generates its response based on the retrieved web pages.
Full entry → - AI Overviews
- AI-generated summary answers displayed directly at the top of search engine results pages, such as Google's AI Overviews.
Full entry → - Zero-Click Search
- A search where users get their answer directly on the results page or within the AI response without clicking through to any website.
Full entry → - AI Visibility Index
- A composite score used by Starbox GEO to measure a brand's exposure in AI responses. Calculated using the same fixed prompt set, it enables direct comparisons between diagnostic rounds.
Full entry → - Fixed Prompt Set
- A set of questions that simulate real inquiries from target buyers. Once finalized during the initial audit, the set remains unchanged, and every retest uses the exact same questions to ensure results are directly comparable.
Full entry → - Retesting
- Rerunning the same fixed question set across AI engines at regular intervals (Starbox GEO defaults to 30 days) to compare indices and metrics against the previous round and verify if the optimization direction is working.
Full entry → - Citation Sources
- The webpages, articles, or platforms referenced and cited by AI when answering questions. Becoming a cited source is a prerequisite for a brand to be recommended by AI.
Full entry → - Entity Consistency
- Keeping basic information like brand name, description, products, and address consistent across your website, wikis, media, and social platforms ensures AI models identify them as the same brand.
Full entry → - E-E-A-T
- Google's four dimensions for evaluating content quality—Experience, Expertise, Authoritativeness, and Trustworthiness (E-E-A-T)—are also widely used to judge whether content is worth citing by AI.
Full entry → - llms.txt
- An explanatory file placed in the website root directory for AI to read, using brief text and links to explain what the site is and which pages matter most.
Full entry → - Structured Data (Schema)
- Use schema.org standards to annotate organization, article, FAQ, and product data within webpages, so search engines and AI can understand the content without guessing.
Full entry → - AI Crawlers
- Automated programs used by AI companies to scrape web content, which may be used to train models or power real-time web retrieval in responses.
Full entry →
Marketing Terms
Content Marketing
Help customers know and trust you through content: topic selection, writing, distribution, and compliance.
- Content Marketing
- A marketing approach that consistently publishes valuable content for target customers to attract attention, build trust, and ultimately drive inquiries and purchases.
Full entry → - Brand Knowledge Base
- Organizing brand product details, target audiences, writing guidelines, case studies, and prohibited terms into an AI-readable repository, so AI pulls facts solely from here when generating content.
Full entry → - Topic Library
- A pre-curated collection of planned topics specifying the question to answer, target audience, and angle, ready to be pulled sequentially or as needed during writing.
Full entry → - Content Matrix
- A coordinated content strategy structured across platforms, accounts, and content types, giving each account a distinct role to publish.
Full entry → - Brand Voice
- Content published from the brand's or founder's perspective, posted on the brand's own accounts.
Full entry → - Third-Party Voice
- Content written from a third-party perspective, such as creators and industry experts, ideally published by third-party accounts, with the brand name and verifiable brand facts appearing naturally in the copy.
Full entry → - Word of Mouth
- Content written in the voice of everyday users and real customers, used to showcase customer case studies and reviews.
Full entry → - Search intent
- The underlying problem a user genuinely wants to solve when searching or asking a question. Common categories include finding information, comparing options, preparing to buy, or navigating to a specific website.
Full entry → - Long-tail keywords
- Longer, more specific search queries. While individual search volume is low, their cumulative volume is massive, user intent is clear, and competition is lower.
Full entry → - Content homogenization
- When content across different brands and articles is virtually identical, readers cannot tell them apart, and search engines and AI models will not prioritize citing them.
Full entry → - Content framework
- A reusable structural pattern for a specific category of content—such as "Problem – Cause – Solution – Case Study"—used to ensure completeness and readability.
Full entry → - Multi-Platform Distribution
- Adapting content on a single topic to fit each platform's format and audience habits, then publishing it across multiple channels.
Full entry → - Content Calendar
- A scheduled publishing plan by date, detailing which topic goes on which platform each day and who owns it.
Full entry → - UGC (User-Generated Content)
- Content created and shared voluntarily by everyday users, including reviews, unboxings, user experiences, and Q&A replies.
Full entry → - KOL / KOC
- KOLs (Key Opinion Leaders) are influencers or experts with authority in a niche; KOCs (Key Opinion Consumers) are everyday users with smaller followings who genuinely use the product and influence those around them.
Full entry → - Advertising Law Compliance
- Marketing content complies with regulations such as the Advertising Law of the People's Republic of China, avoiding superlative claims, false advertising, and regulatory overreach in sensitive industries.
Full entry →
Customer Acquisition and Conversion
From visibility to closed deals: how to measure leads, conversion, costs, and repeat purchases.
- Customer Journey
- The complete path a customer takes from first discovering a brand to researching, comparing, purchasing, using, repurchasing, and recommending it.
Full entry → - Marketing Funnel
- Dividing prospects from initial exposure to closed deal into stages where audience size narrows at each step, helping identify where drop-offs are highest.
Full entry → - Sales Leads
- Potential customers who have expressed purchase intent and provided contact information, serving as the starting point for sales follow-up.
Full entry → - Conversion Rate
- The percentage of total users who complete a target action, such as the proportion of visitors who leave contact details or inquiries that close into deals.
Full entry → - Customer Acquisition Cost (CAC)
- The average cost to acquire a new customer, calculated by dividing total sales and marketing spend over a given period by the number of new customers acquired.
Full entry → - Customer Lifetime Value (LTV)
- The total revenue or profit a brand expects to generate from a customer throughout their business relationship.
Full entry → - Return on Investment (ROI)
- The ratio of gain generated by an investment relative to its cost, used to evaluate whether an investment is worthwhile.
Full entry → - ROAS (Return on Ad Spend)
- Sales generated by advertising divided by advertising spend, used to measure campaign effectiveness.
Full entry → - Landing page
- The page users land on after clicking an ad or link, purpose-built for a single conversion goal, such as booking an appointment, signing up, downloading, or purchasing.
Full entry → - A/B testing
- Showing two versions simultaneously to two similar user groups to see which one performs better.
Full entry → - Private domain traffic
- A customer audience that a brand can repeatedly reach directly without paying per touchpoint, such as WeCom contacts, community groups, and WeChat Official Account followers.
Full entry → - Public Traffic
- Traffic accessible to everyone across platforms—such as search, recommendation feeds, and AI search—which brands acquire through content or advertising.
Full entry → - Product Seeding
- Generating interest and purchase intent among potential customers through authentic experience sharing.
Full entry → - CRM (Customer Relationship Management)
- Methods and systems for recording and managing customer information, follow-up processes, and closed deals.
Full entry → - Repeat Purchase Rate
- The percentage of total purchasing customers who buy again within a given timeframe.
Full entry →
