FREE LEARNING TOOL
128 terms · 8 categories · EN / 中文
Understand essential AI, marketing, search and automation terms through simple explanations and practical examples.
Confused by terms such as API, MCP, SEO, AEO and GEO? Search or browse the essential language of AI marketing, explained clearly with practical examples.
Every definition is free to read. No sign-up required.
128 terms available
Start with ten ideas you will meet again and again.
Computer systems that perform tasks such as recognising patterns, generating content or making predictions.
Learn more PromptingThe instruction or input given to an AI system to guide its response.
Learn more Automation and IntegrationA defined way for software systems to request data or actions from each other.
Learn more Automation and IntegrationAn open protocol that standardises how AI applications connect to external tools, resources and context.
Learn more Search and AI VisibilityImproving a website so search engines can discover, understand and present useful pages.
Learn more Search and AI VisibilityOrganising content so systems can identify clear, useful answers to questions.
Learn more Search and AI VisibilityImproving content and entity clarity so generative systems can understand and potentially reference it.
Learn more Search and AI VisibilityPractices intended to improve how large language models understand and represent a brand, entity or content.
Learn more AI FoundationsAn AI system that can select tools and take steps towards a goal within assigned permissions.
Learn more Trust and Responsible AIDeveloping and using AI with attention to safety, fairness, privacy, accountability and human needs.
Learn moreSEO helps content appear in traditional search. AEO makes clear answers easier to identify. GEO improves how generative AI systems can understand and potentially reference content. LLMO focuses on how clearly large language models understand a brand, entity and its content. These practices overlap; they are not completely separate systems, and none guarantees rankings or AI citations.
YOUR REFERENCE LIBRARY
A short meaning first. Open any term for examples, actions and common mistakes.
No matching term found. Try a different keyword or browse the categories below.
A/B Testing
Comparing randomly assigned versions of an experience to evaluate a defined outcome.
It can provide stronger evidence about a change's effect.
Visitors are randomly shown one of two enquiry button labels.
Set a primary metric, randomise fairly and allow sufficient data before interpreting.
Stopping as soon as a preferred version appears ahead.
Answer Engine Optimisation
Organising content so systems can identify clear, useful answers to questions.
It improves answer clarity across search and assistant experiences.
A service page answers a genuine pricing question with scope and conditions.
Use question-led headings where natural and answer directly with supporting detail.
Assuming FAQ markup guarantees an answer placement.
Agentic Artificial Intelligence
An approach in which AI plans or adjusts multi-step actions to pursue a goal.
It is useful when the next step depends on previous results.
A research workflow revises its queries when initial evidence is weak.
Define stopping rules, budgets and points for human review.
Calling every fixed chatbot script an autonomous agentic system.
Artificial General Intelligence
AI with broad capabilities to learn and perform many different cognitive tasks at a human-comparable level. Definitions and evaluation standards vary.
AGI describes the breadth of intelligence, while generative AI describes the ability to create content. The distinction helps teams assess product claims realistically.
When a supplier calls its marketing assistant AGI, a team checks its performance on unfamiliar briefs, research, planning and analysis rather than judging it by one polished advert.
Evaluate tools against your actual tasks, check evidence and limitations, and retain human review. Do not base business plans on a promised AGI arrival date.
Assuming that fluent conversation, content generation or an autonomous AI agent is sufficient evidence of AGI. AGI also does not automatically imply consciousness.
Artificial Intelligence
Computer systems that perform tasks such as recognising patterns, generating content or making predictions.
AI can support different tasks; not every AI tool generates content.
A retailer predicts demand and separately uses AI to draft campaign copy.
Choose a specific task and compare results with a human baseline.
Treating every automated rule as intelligent, or assuming AI understands like a person.
Artificial Intelligence Agent
An AI system that can select tools and take steps towards a goal within assigned permissions.
It can coordinate tasks, but its actions may affect real systems.
An agent drafts a campaign report using approved analytics tools.
Limit permissions and require approval before publishing or spending.
Giving an agent unrestricted access because its first draft looked good.
Artificial Intelligence Avatar
A digital character used to represent a person or presenter with AI assistance.
It can support repeatable presentation formats.
An authorised virtual presenter introduces a training module.
Obtain likeness and voice permissions and disclose synthetic presentation where appropriate.
Using someone's likeness without permission.
AI Governance
The policies, roles and processes used to direct and oversee AI use.
It clarifies who can approve, monitor and respond to AI-related risks.
A company maintains approved tools and a review process for new uses.
Document ownership, acceptable uses and incident handling.
Writing a policy without putting review and accountability into practice.
Low-Quality AI-Generated Content
A critical label for low-value, inaccurate or repetitive AI-generated content.
Volume alone can weaken audience trust and usefulness.
A feed repeats generic AI posts with unsupported claims.
Publish only content with a clear audience purpose and editorial review.
Assuming more generated content means better marketing.
AI-Generated UGC-Style Content
AI-produced content designed to resemble informal user-created media.
It offers a production style, but is not evidence of a real customer's experience.
A fictional AI presenter demonstrates a product using approved claims.
Label synthetic presenters appropriately and avoid invented testimonials.
Calling a fictional AI review a genuine customer review.
AI Visibility
The extent to which an entity or content appears accurately in AI-generated responses within a defined set of queries.
It helps assess discoverability beyond traditional search results.
A team records whether its brand appears across a repeatable set of relevant questions.
Define queries, platforms, dates and evaluation criteria before comparing results.
Treating one response or a technical checklist score as universal AI visibility.
Application Programming Interface
A defined way for software systems to request data or actions from each other.
It connects tools without manual copying.
A website sends an approved enquiry into a CRM through its API.
Read the interface contract, protect credentials and handle errors.
Exposing secret keys in browser code or assuming every request succeeds.
Marketing Attribution
Assigning credit for an outcome to marketing touchpoints using a defined model.
It shapes how channel performance is interpreted.
A last-click report credits the final tracked channel before an enquiry.
State the model, lookback window and tracking limitations.
Confusing assigned credit with proof that a channel caused the outcome.
Automation
Using systems to perform defined tasks with reduced manual intervention.
It can save repetitive effort and improve consistency.
A form submission creates a task for the sales team.
Start with a stable process and add monitoring and recovery.
Leaving automated failures unnoticed or removing necessary review.
Brand Awareness
The extent to which people recognise or recall a brand or offer.
People usually need to know an option exists before considering it.
A useful introductory video reaches a relevant new audience.
Choose a clear message and measure recognition where feasible.
Equating impressions alone with remembered brand awareness.
Inbound Link
A link from another website to your website.
Relevant editorial links can help discovery and indicate useful references.
An industry publication links to a genuinely useful guide.
Create reference-worthy content and build legitimate relationships.
Buying manipulative links or judging quality only by link count.
AI Bias
Systematic tendencies in data, design or outputs that can produce skewed or unfair results.
It can affect who receives useful or harmful outcomes.
A tool performs worse on one language variety than another.
Test relevant groups and contexts, inspect errors and improve the process.
Assuming a large model is automatically neutral.
Brand Safety
Practices that reduce harmful or unsuitable associations involving a brand's content and advertising.
It helps protect trust and reputation.
A team checks generated imagery and ad placements before a campaign.
Define unsuitable content, review claims and maintain an incident response process.
Assuming automated filters catch every reputational risk.
Customer Acquisition Cost
The cost of acquiring customers, calculated using a stated cost scope and customer cohort.
It helps assess acquisition economics.
A business divides included acquisition costs by newly acquired customers for a matched period.
State included costs and account for the sales cycle when matching customers.
Using only ad spend while labelling the result fully loaded CAC.
Causation
A relationship in which changing one factor produces a change in another under stated conditions.
It matters when deciding which intervention will help.
A well-designed randomised test estimates whether a page change improves enquiries.
Use suitable experiments or careful causal methods and state limitations.
Claiming causation from a simple before-and-after comparison.
Chain-of-Thought Reasoning
Intermediate reasoning steps used while working towards an answer.
Breaking a problem down can help, but displayed steps are not proof of correctness.
A planner presents assumptions and a concise rationale for a channel choice.
Ask for useful explanations, evidence and checks; follow the model's guidance on reasoning prompts.
Demanding hidden internal reasoning or trusting a long explanation without verification.
Character Consistency
Keeping a character's recognisable features stable across generated images or scenes.
It helps audiences recognise the same character.
A mascot keeps its colours and facial features across campaign images.
Maintain approved references and compare every output.
Assuming the same prompt guarantees identical appearance.
Chatbot
Software that interacts through conversation, using rules, AI or both.
It can help people find information or complete simple tasks.
A website bot answers basic service questions and offers human contact.
Define supported tasks, reliable sources and escalation routes.
Letting the bot invent commitments or hide its limitations.
Source Citation
A reference identifying the source used to support a statement.
It helps readers trace evidence.
An AI answer links to the original report behind a claim.
Open the cited source and verify that it supports the specific statement.
Assuming a plausible-looking citation exists or proves the claim.
Consent
An informed agreement to a specified action or use, with legal requirements depending on context and jurisdiction.
It respects people's choices about participation and information use.
A subscriber actively chooses to receive a particular newsletter.
Explain the purpose clearly, record choices and honour withdrawal where applicable.
Using vague blanket permission or assuming silence means agreement.
Purchase Consideration
The stage in which someone evaluates whether an offer fits their needs.
Useful evidence can reduce uncertainty.
A prospect compares service scope, examples and pricing.
Provide honest comparisons, limitations and answers to common questions.
Applying purchase pressure before answering genuine concerns.
Constraint
A boundary that the output or process must respect.
It makes practical requirements testable.
A social caption must fit the platform limit and avoid unapproved claims.
List measurable limits and check the final output against them.
Assuming AI always obeys a stated word count or exclusion.
Content Calendar
A schedule of planned content, channels, owners and publication dates.
It coordinates production and review.
A team tracks next month's posts and approval deadlines.
Include purpose, owner, status and review time.
Filling every date without a clear audience benefit.
Content Credentials
Provenance information that can record how digital media was created or edited using supported standards.
It adds context about a file's history, not proof that every claim is true.
An image carries a signed record of its creation and edits.
Use compatible tools and inspect the available provenance alongside other evidence.
Treating a credential as proof of truth or its absence as proof of fraud.
Content Pillar
A core subject area used to organise a brand's content planning.
It supports a coherent, repeatable editorial direction.
A retailer plans content around product care, selection and customer use.
Choose topics that connect audience needs with your expertise.
Choosing broad themes that do not guide useful content decisions.
Context
Information supplied to help AI interpret a task, audience or situation.
Useful context makes generic output more relevant.
A brief includes brand tone, offer details and the intended audience.
Include relevant, approved facts and remove sensitive data that is unnecessary.
Pasting every available document without identifying what matters.
Context Window
The amount of content a model can handle in one request, subject to model limits.
Long chats and documents may crowd out relevant information.
A team provides a brand guide, campaign brief and examples in one request.
Prioritise relevant context and check input and output limits.
Assuming everything in a long chat will always be retained and used accurately.
Conversion
A defined action that counts as progress towards a business or user goal.
It makes campaign outcomes measurable.
A completed enquiry form is counted as a lead conversion.
Specify the action and validate its tracking.
Counting all clicks as meaningful business results.
Conversion Rate
The percentage of a defined audience or interactions that completes a specified conversion.
It helps compare how effectively a flow supports its goal.
Ten enquiries from 200 eligible visits gives a 5% visit-to-enquiry rate.
Define numerator, denominator and time period before comparing.
Comparing rates that use different conversion definitions or denominators.
Copyright
Legal rights concerning certain original creative works, with protection and exceptions varying by jurisdiction.
It affects what content you can copy, adapt or publish.
A marketer checks permission before reusing a photograph.
Verify licences, ownership and local rules; seek qualified advice for uncertain commercial uses.
Assuming online availability or AI generation makes all content free to use.
Correlation
A statistical association in how variables vary together.
It helps identify patterns worth investigating.
Sales and ad spend rise together during a festive period.
Check alternative explanations and the data's scope.
Concluding that one variable caused the other merely because they moved together.
Crawlability
Whether a crawler can discover and fetch a page's content.
Content cannot be assessed if access is blocked or broken.
A team finds that important pages return errors to crawlers.
Check HTTP responses, internal links, robots rules and accessible HTML.
Assuming crawlable means indexed or ranked.
Customer Relationship Management
Practices and systems for organising customer relationships, interactions and related information.
It helps teams follow up consistently.
A sales team records enquiry status and agreed next steps.
Keep records accurate, restrict access and respect contact preferences.
Treating a CRM as permission to send messages to every stored contact.
Call to Action
A prompt inviting someone to take a specific next step.
It makes the intended action clear.
A button says Book a Consultation.
Use a specific verb and explain what happens next.
Using misleading labels or competing primary actions.
Click-Through Rate
The percentage of measured impressions that produces the defined click action.
It indicates how often displayed content attracts clicks.
An ad with 20 qualifying clicks from 1,000 impressions has a 2% CTR.
Define the click type and check tracking and reporting scope.
Treating clicks as proof of sales or satisfied visitors.
Customer Journey
The steps and interactions a person experiences before, during and after a purchase.
It reveals where information or support is needed.
A buyer discovers a service, compares options and contacts support after purchase.
Map real touchpoints and identify friction with customer feedback.
Assuming everyone follows the same linear path.
Customer Persona
A research-informed description of a representative customer type.
It makes audience needs easier to discuss.
A team summarises a buyer's tasks, concerns and purchase criteria.
Use evidence from interviews and behaviour, then revisit assumptions.
Inventing detailed personal traits without supporting research.
Data Privacy
The appropriate handling of information about people, including collection, use, access and retention.
It protects people's interests and supports trust.
A team removes unnecessary personal details before using an AI tool.
Check applicable local rules, minimise data and seek qualified advice for compliance decisions.
Assuming one country's rules apply everywhere or every AI tool is suitable for personal data.
Data Visualisation
Representing data visually to make patterns and comparisons easier to understand.
It supports clearer decisions and communication.
A line chart shows monthly enquiry trends.
Choose an appropriate chart and label units, periods and sources.
Using distorted scales or omitting relevant context.
Deepfake
Synthetic or manipulated media that convincingly imitates a person's likeness or voice.
It can mislead people about who said or did something.
A fabricated clip appears to show an executive announcing a policy.
Verify important claims through trusted channels and obtain permission for legitimate simulations.
Sharing a convincing clip before verifying its origin and context.
Digital Human
A digitally created human-like figure, sometimes animated or interactive with AI.
It combines visual identity with presentation or interaction.
A virtual guide answers basic exhibition questions.
Define its purpose, limitations and human escalation route.
Implying the character is a real human employee.
Digital Public Relations
Building a public reputation through online coverage, relationships and useful stories.
It can bring credible attention and relevant references.
A business shares original research with a relevant publication.
Offer accurate, newsworthy material and disclose commercial relationships.
Treating paid placement as independent editorial endorsement.
Experience, Expertise, Authoritativeness and Trustworthiness
A framework used in Google's search quality evaluation guidance to assess aspects of content quality.
It encourages evidence of relevant experience, knowledge and trust.
A guide identifies its author and explains the basis of its claims.
Show accurate authorship, useful evidence and clear correction channels.
Treating E-E-A-T as a single measurable website score or adding fake credentials.
Engagement Rate
A ratio of defined interactions to a specified base such as reach or impressions.
It helps assess response relative to exposure.
A team divides selected post interactions by reach and states that definition.
Specify which actions count and which denominator you use.
Comparing rates with different interaction or denominator definitions.
Entity-Based Search Engine Optimisation
Clarifying the people, organisations, products or concepts described by a website and their relationships.
It helps distinguish similar names and connect supporting information.
A consultant uses consistent identity details and verified profile links.
Create a clear about page and accurate identity markup with genuine references.
Inventing credentials or external profiles to strengthen an entity.
AI Explainability
The ability to provide useful, understandable reasons or descriptions of how a system reached an outcome.
It supports review and appropriate reliance.
A recommendation tool shows relevant factors behind a suggestion.
Use explanations suited to the audience and validate them against system behaviour.
Mistaking a fluent AI explanation for a faithful account of its internal reasoning.
Fact-Checking
Checking whether specific factual claims are supported by reliable evidence.
It reduces avoidable inaccuracies before publication.
An editor verifies a quoted figure against its original report.
Break content into checkable claims and record sources, dates and uncertainty.
Asking the same AI to confirm itself and treating that as independent evidence.
Few-shot Prompting
Giving a small set of examples in a prompt to demonstrate the desired pattern.
Examples communicate expectations that abstract instructions may miss.
A brand supplies two approved captions before asking for another.
Choose varied, accurate examples and label inputs and outputs clearly.
Using examples that contradict the instructions or include private customer data.
Fine-tuning
Additional training that changes a model's parameters using selected examples.
It can adapt behaviour or output style for a recurring task.
A team trains a model to classify enquiries using carefully labelled examples.
First establish a baseline, then evaluate on examples kept out of training.
Using fine-tuning as the default way to keep changing facts up to date.
First-Party Data
Data an organisation collects directly through its own relationships and interactions.
It can support relevant service and measurement with appropriate controls.
Customers choose topics when subscribing to a company's newsletter.
Explain purposes, collect only what is needed and protect access.
Assuming directly collected data can be reused for any purpose.
Foundation Model
A broadly trained model that can be adapted or prompted for many downstream tasks.
One model may support several workflows rather than one narrow function.
A business uses the same base model for drafting, summarising and classification.
Evaluate each intended task separately before choosing the model.
Assuming broad training means equal competence in every domain.
Generative Artificial Intelligence
AI that generates text, images, audio or other content from learned patterns.
It speeds up drafts and exploration, but output still needs review.
A marketer creates three campaign concepts from an approved brief.
Supply context, compare alternatives and verify claims before publication.
Assuming newly generated content is automatically true or original.
Generative Engine Optimisation
Improving content and entity clarity so generative systems can understand and potentially reference it.
It supports accurate interpretation when AI systems retrieve or discuss information.
A business publishes specific service facts, authorship and supporting sources.
Make claims clear, verifiable and accessible; monitor relevant queries over time.
Promising AI citations or treating GEO as a replacement for SEO.
AI Hallucination
An AI output that appears plausible but is false, unsupported or inconsistent with the source.
Confident mistakes can damage customer trust.
A draft invents a product certification that the brand never received.
Check claims and citations against original sources.
Asking the same model whether it is sure and treating yes as verification.
Human Oversight
Human responsibility for monitoring, directing and correcting a system across its use.
It provides accountability beyond a single approval step.
An owner monitors an AI assistant's errors and can pause it.
Assign competent owners and practical intervention powers.
Naming an owner who cannot inspect or stop the system.
Human-in-the-Loop
A design in which a person reviews or intervenes at defined points in an automated process.
It adds judgement where errors or consequences matter.
AI drafts a customer reply and a staff member approves it before sending.
Give reviewers evidence, time and a genuine ability to reject.
Adding a rubber-stamp approval step with no meaningful review.
Image-to-Image Generation
Generating a new image using an existing image as input.
It helps explore variations with visual guidance.
A designer tests backgrounds around a supplied product photo.
Use an authorised reference and check product fidelity.
Assuming the original subject will remain unchanged.
Image-to-Video Generation
Generating motion from an input image.
It brings approved visuals into video concepts.
A product still becomes a gentle camera move.
Use a suitable source image and specify restrained motion.
Ignoring product distortion during movement.
Impression
A counted instance of content or an advertisement being displayed under a platform's measurement rules.
It indicates delivery volume.
An advertisement records several displays to the same person.
Check the platform definition and compare like-for-like periods.
Assuming every impression was noticed or came from a different person.
Indexability
Whether a page is eligible to be included in a search engine's index.
It affects the possibility of appearing in search results.
An accidental noindex tag excludes a useful public page.
Review noindex directives, canonical choices and index status where available.
Expecting every eligible page to be indexed immediately.
Image Inpainting
Replacing or generating content within a selected image region.
It allows focused edits without rebuilding the whole image.
A designer replaces a distracting background object.
Mask only the intended region and inspect neighbouring details.
Overlooking changes outside the intended edit.
Software Integration
Connecting systems so they exchange information or coordinate actions.
It reduces disconnected work and duplicate entry.
An enquiry form passes relevant details into a customer system.
Define field mappings, permissions and failure handling.
Connecting systems without checking data quality or ownership.
Intellectual Property
A group of legal protections relating to creations and identifiers, such as copyright, trademarks and patents.
Different rights can apply to the same campaign asset.
A product visual may involve artwork copyright and brand trademarks.
Identify relevant rights and licences in the markets where you will use the asset.
Treating all intellectual property as the same right; get professional advice when needed.
Iterative Prompting
Improving output through successive rounds of specific feedback.
Targeted revision is more controlled than repeatedly starting over.
A marketer keeps an email structure but asks to simplify its opening.
Change one clear issue at a time and preserve approved elements.
Giving conflicting corrections without restating the current goal.
Search Keyword
A word or phrase used to describe a topic or search query.
It helps connect audience language with relevant content.
A café notices customers search for meeting-friendly cafés.
Use natural audience wording in an informative page.
Repeating an exact phrase excessively to chase rankings.
Knowledge Base
An organised collection of information used for reference and support.
It gives people and tools a shared source of approved information.
A team maintains current product instructions and support answers.
Assign owners, review dates and access levels.
Feeding outdated or contradictory documents into an AI assistant.
Key Performance Indicator
A selected measure used to assess progress towards an important objective.
It focuses attention on outcomes that matter.
A support team tracks resolved enquiries rather than only message volume.
Choose a metric tied to a goal and define how it is calculated.
Calling every available metric a KPI.
Landing Page
A page designed to receive visitors and guide them towards a focused goal.
It connects a campaign promise with a clear next step.
An ad links to a page explaining the exact advertised workshop.
Align the headline, offer, evidence and action with the referring message.
Sending visitors to a page that does not match the ad.
Lead Generation
Attracting and collecting interest from people who may become customers.
It starts a permission-based sales conversation.
A visitor requests a relevant consultation through a clear form.
Ask only for needed information and explain follow-up expectations.
Collecting contacts without a clear purpose or appropriate permission.
Large Language Model
A model trained on large amounts of data to process and generate language.
It helps with writing and analysis but can produce unsupported claims.
A team uses an LLM to turn approved product facts into email drafts.
Provide reliable facts and check the output against them.
Treating a fluent answer as evidence of accuracy.
Large Language Model Optimisation
Practices intended to improve how large language models understand and represent a brand, entity or content.
Clear, consistent information can reduce ambiguity across sources.
A company aligns its name, service descriptions and verified profiles.
Publish consistent facts and test relevant prompts while recording dates and context.
Assuming a public webpage change immediately alters a model's stored knowledge.
Local Search Engine Optimisation
Improving discoverability for searches with local intent.
It helps nearby customers find relevant businesses and services.
A shop maintains accurate location, hours and local service information.
Keep real business details consistent and provide useful location content.
Creating misleading locations or unsupported service-area claims.
Low-Code Development
Building applications with visual components plus selected custom code.
It balances rapid assembly with custom logic.
A team adds a custom calculation to a visual workflow.
Document custom code and test how it interacts with the platform.
Leaving custom logic unowned because most of the app is visual.
Machine Learning
Methods that learn patterns from data to make predictions or decisions.
Useful predictions depend on relevant data and suitable evaluation.
A campaign model estimates which leads may respond to an offer.
Separate training data from evaluation data and monitor performance changes.
Assuming a pattern learned last year will remain reliable forever.
Marketing Automation
Using software to coordinate marketing tasks based on defined rules and relevant customer events.
It can deliver timely follow-up at a manageable scale.
A requested resource is followed by an optional, consented learning sequence.
Map permissions, content rules, unsubscribe handling and measurement.
Continuing messages after someone opts out.
Marketing Funnel
A simplified model of movement from initial awareness towards a desired action.
It helps plan messages and measure stage-specific progress.
A campaign uses introductions for new audiences and demonstrations for interested buyers.
Define useful stages and measures for your business.
Treating the model as a literal path every customer follows.
Model Context Protocol
An open protocol that standardises how AI applications connect to external tools, resources and context.
It gives compatible AI tools a shared connection pattern.
An AI assistant accesses an approved knowledge tool through an MCP server.
Review server permissions and data access before connecting; enable only the tools needed for the task.
Confusing MCP with MPC or assuming a connection automatically makes every tool safe.
Multimodal Artificial Intelligence
AI that works with more than one form of input or output, such as text and images.
It lets teams discuss visual work alongside written instructions.
A designer uploads a poster and asks for a readability review.
Check which media the tool supports and inspect its interpretation.
Assuming every multimodal model can create and understand every media type.
Negative Prompt
Instructions describing elements to exclude, often in image generation.
It can reduce unwanted visual features when the tool supports it.
A designer asks an image model to avoid text overlays and watermarks.
Use the tool's supported controls and inspect the actual result.
Assuming negative prompts work identically across all image tools.
No-Code Development
Building software behaviour mainly through visual tools and configuration.
It lets more people create useful internal tools.
An operations team builds an approval form using a visual platform.
Check access controls, limits and how to export your data.
Assuming no code means no technical risks or maintenance.
Omnichannel Marketing
Coordinating channels so customers experience a coherent journey across them.
It reduces conflicting information and repeated effort.
A customer sees consistent offer terms online and in store.
Align service information and handovers while respecting data permissions.
Assuming being present on many channels means they work together.
Image Outpainting
Generating new content beyond an image's original boundaries.
It helps adapt an image to different layouts.
A square visual is extended to fit a wide banner.
Plan the crop and inspect perspective and lighting at the joins.
Assuming invented surroundings represent the real location.
Output Format
The structure requested for a response, such as a table, list or JSON.
Consistent structure makes output easier to review or use in tools.
A campaign plan is returned with columns for channel, message and owner.
Provide a schema or example and validate machine-readable output.
Assuming JSON-looking text is valid JSON or factually correct.
Personal Data
Information that identifies or can relate to an identifiable person, subject to applicable legal definitions.
It requires appropriate handling throughout a workflow.
An enquiry record contains a name, email address and message.
Identify what you collect, why it is needed and who can access it; check local requirements.
Assuming removing a name always makes a record anonymous.
Marketing Personalisation
Adapting an experience or message using relevant information about a person or segment.
It can make communication more useful when done appropriately.
A subscriber receives content matching topics they selected.
Use appropriate data, offer controls and test usefulness.
Inferring sensitive traits or making people feel secretly monitored.
Brand Positioning
The place an offer aims to occupy in a particular audience's mind relative to alternatives.
It guides what the brand emphasises.
A software company focuses on simplicity for small teams.
Choose a relevant difference that the experience can support.
Claiming to be the best choice for every audience.
Predictive Analytics
Using data and models to estimate future or unknown outcomes.
It supports planning under uncertainty.
A retailer estimates demand using historical sales and relevant conditions.
Validate on unseen data and monitor changing conditions.
Treating a prediction as certainty or ignoring biased inputs.
Prompt
The instruction or input given to an AI system to guide its response.
Clear direction reduces avoidable revisions.
A marketer requests three email subject lines for existing customers.
Specify the task, audience, useful context and desired output.
Giving vague instructions such as make it better without explaining the goal.
Prompt Engineering
Systematically designing and testing instructions to improve results on a task.
Reusable prompts help teams work consistently.
A team tests a product-description prompt on several product types.
Keep test examples and revise based on observed failures.
Searching for a magic phrase instead of testing the workflow.
Prompt Injection
Untrusted input that tries to redirect an AI system away from its intended instructions.
External documents can carry malicious instructions as well as useful facts.
A retrieved webpage tells an assistant to reveal internal customer records.
Treat retrieved text as data, limit tool permissions and require approval for sensitive actions.
Assuming instructions inside a webpage are authorised by the user.
Retrieval-Augmented Generation
A method that retrieves relevant information and provides it to a model when generating an answer.
It can ground responses in current or private material without retraining the model.
A product assistant retrieves approved specifications before answering.
Maintain sources, access controls and tests for unsupported answers.
Assuming retrieval eliminates hallucinations or permission risks.
Audience Reach
The estimated number of distinct people or accounts exposed to content in a defined period.
It helps distinguish audience breadth from repeated displays.
A post has more impressions than reached accounts because some saw it twice.
Record the platform's deduplication method and reporting period.
Adding platform reach figures as if audiences never overlap.
Reasoning Model
A model designed to spend additional computation on working through complex tasks before answering.
It may help with multi-step planning, at the cost of latency or usage.
A marketer compares a launch plan against budget and scheduling constraints.
State the goal and constraints clearly; evaluate the final answer and evidence.
Demanding hidden internal reasoning or assuming more computation guarantees correctness.
Responsible Artificial Intelligence
Developing and using AI with attention to safety, fairness, privacy, accountability and human needs.
It makes benefits and possible harms part of everyday decisions.
A team reviews a customer-facing AI tool before release and monitors problems afterward.
Set clear boundaries, assign owners and provide a route to report errors.
Treating responsibility as a one-time checklist.
Advertising Retargeting
Showing advertising to people based on previous relevant interactions, subject to platform rules and applicable consent requirements.
It can reconnect with people who have shown interest.
A visitor sees a reminder about a product they viewed.
Respect consent choices, exclude completed goals where appropriate and limit repetition.
Following people excessively or ignoring privacy choices.
Return on Ad Spend
Revenue attributed to advertising divided by advertising spend.
It provides a revenue-based view of ad efficiency.
RM4,000 attributed revenue from RM1,000 ad spend gives a 4× ROAS.
Check attribution rules and evaluate margins and other costs separately.
Treating revenue return as profit.
Schema.org Markup
Structured data using the Schema.org vocabulary to describe entities and content.
A shared vocabulary makes descriptions more consistent.
A glossary marks visible definitions as DefinedTerm entries.
Choose suitable types, validate syntax and keep the markup aligned with the page.
Expecting valid markup to guarantee rich results or AI citations.
Search Intent
The underlying goal behind a person's search query.
Matching the goal makes content more useful.
Someone searching for a product comparison may need decision criteria rather than a sales slogan.
Identify the task the query suggests and design the page to fulfil it.
Assuming every searcher is ready to buy.
Semantic Search Engine Optimisation
Creating connected content that addresses meaning, context and related concepts rather than isolated keywords.
It supports a more complete answer to a visitor's need.
A buying guide covers use cases, limitations and comparison criteria.
Map the audience's questions and link genuinely related explanations.
Adding every related phrase without improving understanding.
Sentiment Analysis
Estimating expressed attitudes such as positive, negative or neutral from content.
It can organise large volumes of feedback for review.
A team groups comments by likely sentiment before reading samples.
Validate results in the relevant language and inspect sarcasm and mixed opinions.
Assuming sentiment labels reveal a person's true feelings.
Search Engine Optimisation
Improving a website so search engines can discover, understand and present useful pages.
It helps people find relevant content through search.
A retailer improves product pages, internal links and descriptive titles.
Start with crawlable pages, useful content and accurate page metadata.
Promising guaranteed rankings or stuffing pages with keywords.
Social Listening
Monitoring and interpreting relevant public online conversations about topics or brands.
It reveals questions, concerns and emerging themes.
A team notices repeated questions about product instructions.
Review context and combine automated summaries with human reading.
Treating a small vocal sample as representative of all customers.
Source Verification
Checking the identity, relevance, reliability and context of an information source.
A real source may still be outdated or fail to support a claim.
A team checks an official report's date and underlying data.
Find the original source and confirm the exact passage and context.
Assuming a professional-looking page or citation is trustworthy.
Structured Data
Information organised in a predictable machine-readable format.
It helps software interpret explicitly labelled facts.
A webpage describes its author and content type using JSON-LD.
Mark up accurate facts that match visible content.
Using markup to claim content or credentials absent from the page.
Style Reference
A reference used to guide visual qualities such as colour, lighting and texture.
It communicates visual direction more clearly.
A designer uses an authorised mood board to guide a campaign palette.
Describe the qualities you need and check usage rights.
Assuming a style reference authorises copying protected material.
Synthetic Media
Images, audio, video or text created or substantially altered by computational systems.
It changes how content is produced and verified.
A brand uses an AI-generated voice for an approved script.
Check rights, accuracy and disclosure needs before release.
Presenting fabricated events as genuine recordings.
System Prompt
High-level instructions set by an application to guide model behaviour.
It can establish role, boundaries and response rules across interactions.
A support assistant is instructed to use approved product information.
Keep rules explicit and enforce important restrictions in application code.
Treating a system prompt as a reliable security boundary by itself.
Target Audience
The group of people a message or offer is intended to serve.
It guides relevant messaging and channel choices.
A workshop targets marketing managers who manage small teams.
Define shared needs, constraints and decision context.
Targeting everyone with the same message.
Technical Search Engine Optimisation
Improving the technical conditions that let search engines access, interpret and index a website.
Technical barriers can hide otherwise useful content.
A team fixes broken internal links and accidental noindex directives.
Check responses, crawling rules, canonical links and readable initial HTML.
Assuming a technically valid page will rank without useful content.
Text-to-Image Generation
Generating images from written descriptions.
It speeds up visual exploration.
A café explores poster concepts from a campaign brief.
Describe subject, composition and intended use; review the output.
Assuming generated text and product details are accurate.
Text-to-Video Generation
Generating video from a written description.
It supports quick motion concepts.
A team drafts a short scene of a delivery arriving.
Specify one clear scene and review movement frame by frame.
Expecting long, complex scenes to remain consistent automatically.
Token
A unit a model uses to process content; it may be part of a word or another encoded element.
Token counts affect context limits, cost and output length.
A long bilingual product catalogue uses more tokens than a short brief.
Use the tool's token estimate and remove irrelevant material.
Assuming one token always equals one word or one Chinese character.
Topic Cluster
A group of related pages connected around a central subject.
It helps visitors explore a subject in depth.
A marketing guide links to focused pages on email, content and measurement.
Choose a meaningful central topic and link useful supporting pages.
Producing near-duplicate pages just to increase page count.
AI Transparency
Providing understandable information about AI use, relevant limitations and responsibilities.
It helps people make informed choices.
A service explains when a response comes from an automated assistant.
Disclose relevant AI involvement and offer clear support routes.
Burying important limitations in vague or inaccessible language.
Automation Trigger
An event or condition that starts a workflow.
It defines when automation should run.
A newly approved record starts an email preparation task.
Specify the event precisely and test duplicates and timing.
Using a trigger that fires repeatedly for the same intended action.
User-Generated Content
Content created by users or customers rather than the brand itself.
It can reveal real experiences and audience language.
A customer shares a video showing how they use a product.
Ask permission before reuse and clarify any incentive.
Treating a sponsored or fabricated testimonial as an independent customer opinion.
Image Upscaling
Increasing image resolution, sometimes using AI to infer added detail.
It can prepare visuals for larger displays.
A small campaign image is enlarged for a presentation.
Inspect edges, faces and text at the final display size.
Mistaking inferred detail for recovered factual detail.
Value Proposition
A clear explanation of who an offer helps, what benefit it provides and why it is relevant.
It connects product features with customer needs.
A service promises simpler appointment scheduling for busy clinics.
State the audience, benefit and credible reason to believe.
Using vague superiority claims without evidence.
Vector Database
A database designed to store and search numerical representations such as embeddings.
It supports similarity-based retrieval for AI applications.
A support assistant retrieves passages similar in meaning to a question.
Evaluate retrieval quality, permissions and update behaviour.
Assuming the nearest match is necessarily correct or authorised.
Verification Prompt
A prompt that asks AI to inspect output against evidence or explicit criteria.
It helps organise review, but does not replace independent checking.
A team asks AI to flag claims missing from the product brief.
Supply sources and ask for uncertainties and mismatches, then verify them.
Treating an AI self-check as independent proof.
Webhook
An event-driven HTTP notification sent from one system to another.
It can start work when something happens.
A payment service notifies an order system after a payment event.
Verify signatures, handle retries and avoid processing the same event twice.
Trusting every incoming request as an authentic event.
Workflow
An ordered set of tasks and decisions used to complete a process.
It clarifies handovers and responsibilities.
A draft moves through writing, review, approval and publication.
Map steps, owners, exceptions and completion criteria.
Automating an unclear process before understanding it.
Zero-Click Search
A search interaction that ends without a click through to an external website.
Useful answers may be consumed directly on the results interface.
A user reads opening hours in a search result and stops searching.
Measure relevant outcomes beyond visits and keep public facts accurate.
Assuming every impression should produce a website click.
Legal and regulatory requirements differ by jurisdiction. The glossary provides general learning information; seek qualified professional advice for decisions about rights, privacy or compliance.
LEARN. APPLY. CREATE.
Understanding the terminology is only the beginning. Learn how to apply AI to creative work, marketing, content and workplace productivity through practical corporate training.