AI and Digital Marketing Glossary: 100+ Terms

Having more terms does not always increase understanding. Sometimes the team uses the same word in three different meanings and only notices it when a decision goes wrong. This dictionary brings together in one place concepts you often encounter in artificial intelligence, automation, SEO, GEO, analytics, and marketing in simple Azerbaijani.

Definitions are kept short. The goal is not to make you memorize the word; it is to clarify what you are talking about in a document, meeting, or when choosing a tool. A term may take on a different nuance depending on the platform. Therefore, it is still your job to check the changing function and rule from the original source before making a final decision.

How to use the dictionary?

Find the word you are looking for using the browser's page search function.
Compare the concept in the definition with your real scenario.
Recheck the variable platform rule from the original source.
Define the term once in the project document and then use it with the same meaning.

AI and digital terms

Artificial intelligence: The capability of a computer system to perform recognition, prediction, text generation, and decision-support tasks that usually require human thinking.
Machine learning: A method where the program produces results for new inputs by learning relationships from example data without each rule being written separately.
Deep learning: A field of machine learning that learns features from complex data like images, audio, and text through multi-layer neural networks.
Neural network: It is a computational model that processes inputs through weights and recognizes patterns by adjusting those weights during training.
Generative AI: It is an artificial intelligence system capable of creating new text, images, audio, video, or code based on the examples it has learned.
LLM: It is a large language model trained on a vast amount of text that understands and generates text by modeling language sequence.
Token: It is a unit of word, word part, punctuation mark, or symbol used by an AI model when processing text.
Prompt: It is the input instruction to a generative AI system that specifies the task, context, boundary, and expected output format.
Context window: The limit of the total number of tokens that the model can take into account at the same time while generating a response.
Embedding: A representation converted into a vector of numbers to calculate the semantic similarity of text, image, or another object.
Vector: A mathematical representation that expresses the selected features of an object with a sequence of numbers and allows for comparison.
Vector database: A database that stores embeddings and is optimized to quickly find objects that are semantically close to a query.
RAG: A retrieval-augmented generation approach in which the model finds relevant information from an external source and adds it to the prompt before responding.
Fine-tuning: Fine-tuning It is the process of retraining with additional examples to adapt the behavior of a pre-trained model to a specific task or style.
Foundation model: A large base model trained on broad data that can later be adapted to many different tasks.
Transformer: An architecture that operates on relationships in a text sequence using an attention mechanism and forms the basis of modern large language models.
Attention: A mechanism that calculates which parts of the input are more related to each other when the model generates an output.
Inference: The stage in which a trained model receives new input and produces predictions, classifications, or generative responses.
Training: The learning process in which the model parameters are calculated and updated on sample data to reduce error.
Dataset: A collection of text, images, numbers, and other samples systematically gathered for training, validation, testing, or analysis.
Model: A mathematical system that stores the learned relationships and computations to produce an output from input data.
Parameter: An internal numeric value learned by the model during training that determines how it will respond to input.
Temperature: A parameter controlling the diversity in the generative model's selection among probabilities; a higher value usually produces more variable responses.
Top-p: The next token of the model regulates that it is taken only from the most likely choices that reach the given total probability.
Hallucination: It is the case when an AI model generates incorrect or fabricated information that is not present in the source, even though it sounds convincing.
Guardrail: It is the technical rule, filter, and control layer applied to limit risky inputs and outputs of an AI system.
Alignment: It is the direction of aligning the behavior of an AI system with human goals, safety requirements, and accepted values.
Bias: It is the tendency arising from data, selection, or model structure that creates systematic prejudice for certain groups and outcomes.
Explainability: The degree to which a person can understand and evaluate the basis on which the model arrived at a result.
Multimodal AI: An AI approach that accepts and processes multiple types of data such as text, image, voice, and video within the same system.
Computer vision: The field that teaches computers to recognize objects, text, motion, and visual relationships in images and videos.
NLP: The field of natural language processing that deals with computers processing, analyzing, and generating human language.
Speech-to-text: Speech recognition technology that automatically converts spoken voice into written text.
Text-to-speech: It is a technology that converts written text into synthetic human speech and is used in voice interfaces.
OCR: It is a technology that converts printed or handwritten text in images and scanned documents into machine-readable text.
API: It is an interface that allows two programs to exchange data and functions with predefined request and response rules.
Webhook: It is a method by which one system automatically sends an HTTP notification to another when a certain event occurs.
SDK: A code library, tool, example, and set of documentation that makes it easier to build an application for a specific platform.
Agent: It is a system that observes the environment, selects the next step to achieve the goal, and can perform certain actions.
AI agent: It is an AI system that plans using a language model and tools, gathers information, and executes authorized operations in stages.
Workflow: It is a specific sequence of tasks, decision points, responsible persons, and information flow to produce a result.
Automation: It is performing repetitive work with clear rules through software and integrations with minimal manual intervention.
Trigger: It is a condition such as a time, form submission, status change, or another event that starts an automated flow.
Action: Trigger is an action performed by the system after it occurs, such as sending a message, creating a record, or updating information.
Integration: It is the establishment of separate programs exchanging data and operations, working as part of a single process.
CRM: A system that manages customer and potential buyer information, communication, sales stages, and next steps.
ERP: A platform that integrates core operations such as finance, purchasing, inventory, production, and human resources under a single data system.
CDP: A platform that combines first-party customer data from various channels into a single profile and prepares it for activation.
Chatbot: A program that receives user questions via text or voice and responds based on rules, a database, or AI.
Copilot: An AI assistant that provides suggestions, text, and operational support within the software used, without fully replacing human work.
Custom GPT: A ChatGPT-based assistant tailored to a specific use scenario with custom instructions, files, and tools.
Open-source model: A model whose weights or source code are provided under an open license for use, verification, and adaptation.
Proprietary model: A model whose weights and internal structure are kept closed by the owner, typically used through a service or license.
Cloud AI: It is an AI solution where computing and model services run on remote cloud infrastructure and are accessed over the network.
Edge AI: It is an approach where AI computation is executed on a phone, camera, or local device close to the data source.
AGI: It is the still controversial AI concept presumed to be able to generally learn and perform a wide variety of intellectual tasks like a human.
AEO: It is an optimization approach so that the content can be directly chosen as an accurate answer to the user's question by search and answer systems.
GEO: It is an optimization carried out for the correct understanding, use, and referencing of brand and content in generative AI responses.
SEO: Improving technical, content, and trust signals so that the site can be found in search engines for relevant queries.
SERP: The result page where the search engine shows organic results, ads, maps, and other blocks in response to a query.
AI Overview: A Google feature that shows a generative summary based on sources at the top of search results for some queries.
Structured data: Markup that reports objects and their relationships on the page to the search engine in a standard, machine-readable format.
Schema markup: Structured data code describing articles, products, persons, and other objects using the Schema.org vocabulary.
Entity: An object is a person, organization, place, product, or concept that can be individually recognized by the search and knowledge system.
Knowledge graph: A data model that stores objects and the relationships between them in the form of a graph, merging facts in context.
Citation: Reference information that shows which source a claim or answer is based on.
Source attribution: Clearly identifying the author, publication, and original source of the information and linking the result to that source.
Zero-click search: The situation where the user receives the answer on the results page and completes the search without visiting any website.
Search intent: The information, comparison, site visit, or purchase intent that a user wants to obtain by writing a query.
Keyword: A word or phrase that people use in search and that relates the content topic to the query intent.
Topic cluster: A content structure that covers a main topic through a pillar page and more specific articles connected to it via internal links.
Pillar page: The main and comprehensive page that explains a broad topic systematically and provides links to subtopics.
Internal link: A link given from one page to another within the same domain, showing connection to users and search engines.
Backlink: A link from another domain to a page on your site that can create discovery, trust, and traffic signals.
Canonical URL: A signal indicating to the search engine the preferred main URL for indexing identical or very similar content.
Indexing: The process by which a search engine processes a discovered page and includes it in its search index.
Crawling: The process by which a search bot finds pages through links and a sitemap and reads their content and resources.
Sitemap: An XML file that presents important URLs on the site and sometimes update information to search engines.
Robots.txt: It is a root-level rule file that tells search bots which paths of the site they should not crawl.
Core Web Vitals: User experience metrics from Google that measure page loading, visual stability, and responsiveness to interactions.
Conversion: The completion by a user of a predefined valuable action such as inquiry, purchase, registration, or download.
Lead: A potential customer who shows interest in a product or service and creates a future sales opportunity with a contact or behavioral signal.
Funnel: A marketing and sales model that shows the stages a person goes through from initial awareness to evaluation and decision.
Landing page: A landing page is a page designed for a campaign, audience, and goal, directing the user to a specific next action.
CTA: A button, link, or brief call-to-action that tells the user what step to take next.
CRO: The process of optimizing the experience and page so that more users complete the desired action without increasing site traffic.
A/B test: A method of comparing the impact of two variants on the same key metric by splitting the audience in a controlled manner.
Retargeting: An approach to show relevant ads again to people who have previously interacted with the site or content.
Attribution: Conversion is a model that shows how revenue is distributed between the channels and campaigns the user interacts with.
KPI: A key performance indicator that measures progress toward the goal, influences decisions, and has a pre-agreed definition.
ROI: A profitability metric that shows the ratio of net profit obtained to the investment made.
CAC: Customer acquisition cost calculated by dividing sales and marketing expenses over a certain period by the number of new customers gained.
LTV: The expected total value or contribution margin a customer will generate throughout their relationship with the business.
Churn: The indicator showing the loss in terms of customers, users, or revenue share who leave the service in a certain period.
Retention: An indicator showing the continuation of product usage or purchasing behavior by customers and users after a certain period.
Personalization: The adaptation of a message, offer, and experience based on existing information and behavioral signals about a person.
Segmentation: The division of an audience into separate groups according to needs, behavior, value, and other meaningful criteria.
First-party data: Data collected directly and with consent by a business through its own website, application, sales, and customer relationships.
Cookie: A small piece of data stored in the browser that is used for purposes such as session, preferences, measurement, and advertising.
Consent: Freely given and retractable permission granted by a person, understanding for what purpose their data will be collected and used.
Data privacy: The principle and practice of collecting and protecting personal data legally, transparently, for the intended purpose, and in minimal quantities.
2FA: An account security method that requires a second proof such as a phone, application, or physical key in addition to the password.
Phishing: A fraud attempt that tries to obtain passwords, payment, and other sensitive information from a user by appearing as a trusted organization or person.
Encryption: It is the mathematical conversion of information into a form that can only be read by the party with the appropriate key.
Backup: It is a verified copy kept separate from the main system for recovery in case of data loss or malfunction.
Prompt injection: It is a type of attack where a user or external content tries to make the AI system break previous instructions and behave without authorization.
Human-in-the-loop: It is a control model in the AI process where the human checks the result, handles exceptions, and gives final approval for critical decisions.

When does a definition become useful?

Deciding to know the terms does not automatically improve things. Saying “We are setting up RAG” is easy. It is harder to say which source will be searched, who needs the result, and where a wrong answer will stop. The benefit arises in the answer given to these difficult questions.

When you encounter a new concept, divide it into four parts: input data, work done, output, and responsible person. Then choose a real example. If the word clarifies that example, it is useful. If it only makes the presentation look more technical, it has not solved anything yet.

Where should the source be checked?

Continuation of the topic

A good dictionary does not prolong the conversation.

It helps to use the same word in the same meaning.