AI Words, Simply Explained

Software that performs tasks such as recognizing patterns, interpreting language, or generating content. Different AI systems do different jobs.

Everyday example: A photo app grouping similar pictures and a chatbot drafting a reply both use AI, but their abilities differ.

Keep in mind: Check what this particular tool can do; the label “AI” is not a promise of accuracy.

AI that produces content such as text, images, audio, or code from patterns learned during training and the input it receives.

Everyday example: You ask for a short invitation and receive a new draft.

Keep in mind: A new draft can contain invented details. Compare it with the facts you supplied.

Software you interact with through messages or speech. Some assistants also have tools for working with files or taking actions.

Everyday example: You type a question, read the reply, then ask for a shorter explanation.

Keep in mind: Conversation does not mean human understanding, personal expertise, or permission to act for you.

The instruction, question, or other input you give the AI. An effective request explains the task and relevant limits.

Everyday example: “Write a friendly reply under 80 words. Use only the facts below.”

Keep in mind: Say what a useful result should look like. You do not need special jargon.

Information available to the AI for the task, including relevant instructions, conversation, and supplied material.

Everyday example: For a day-out plan, your time available and preferred pace are useful context.

Keep in mind: Share only the details needed. Restate an important constraint if the answer overlooks it.

A later message that clarifies, corrects, or extends your original request.

Everyday example: “Keep the same facts, but turn that paragraph into three bullet points.”

Keep in mind: Check the revised answer again. A correction in one place can introduce a change elsewhere.

A trained computational system that an AI product uses to process inputs and generate or classify outputs.

Everyday example: One assistant may offer more than one model for different kinds of tasks.

Keep in mind: The product, model, and connected tools are different things. A model name alone does not tell you every available feature.

A kind of model trained on large amounts of language data to work with text and language patterns.

Everyday example: An LLM can help turn rough notes into a readable paragraph.

Keep in mind: Fluent language can still be wrong. Treat factual claims as things to check.

The process of adjusting a model using data. This happens separately from simply reading your request to answer it.

Everyday example: A model may have learned patterns of how invitations are written before you ask for one.

Keep in mind: How a service stores or uses your chats for future training depends on its policy, account, and settings.

A piece of text used by a language model. A token can be a word, part of a word, or punctuation; it is not a fixed number of words.

Everyday example: A long document uses more tokens than a short paragraph.

Keep in mind: A token limit and a word limit are not interchangeable. You rarely need to count tokens for a first exercise.

The limit on how much information a model can work with in one request. Instructions, conversation, supplied material, and response allowance can share that space.

Everyday example: In a long chat, earlier details may be shortened or left out of the information used for the next answer.

Keep in mind: A large limit does not guarantee that every detail was used correctly. Give the relevant passage and restate the task.

A product feature that can retain or draw on details for later chats. Exactly what it remembers and how you control it varies by service.

Everyday example: An assistant might use a saved preference for brief replies in a new conversation.

Keep in mind: Memory, chat history, and model training are different. Check the service’s settings instead of assuming a new chat starts with no prior information.

AI output presented as if it were factual even though it is false or unsupported by the available evidence.

Everyday example: A summary adds a meeting date that never appeared in your notes.

Keep in mind: Ask where the claim comes from, then inspect the original. Confident wording is not evidence.

The original document, webpage, record, or other material used to support a statement.

Everyday example: The manufacturer’s manual is the source for a product’s cleaning instructions.

Keep in mind: Use the right model, edition, date, and relevant passage. A similar-looking document may not answer your question.

A reference or link attached to a claim. It helps you find the material the answer says it used.

Everyday example: An answer points to page 4 of a manual.

Keep in mind: Open it. A citation can be incorrect, unavailable, or unrelated to the claim beside it.

Using specific source material to inform a response instead of relying only on a model’s learned patterns.

Everyday example: “Use only this event notice to list the time and location. Mark missing details unknown.”

Keep in mind: Source material can be incomplete or misunderstood. Check the answer beside the source.

A systematic tendency in data or a system’s behavior that can lead to incomplete, unbalanced, or unfair output.

Everyday example: A brainstorm repeatedly assumes that everyone over 50 wants the same activities.

Keep in mind: Look for missing perspectives and assumptions. Ask for alternatives that fit the person’s actual interests.

The work you do to establish whether a claim or result is correct.

Everyday example: You reopen the original notice to confirm a date, or check a total with a calculator.

Keep in mind: Asking the same chatbot “Are you sure?” is not an independent check.

Sending a file from your device to an app or online service so it can be stored or processed.

Everyday example: You attach a short practice document to a chat.

Keep in mind: Inspect the exact file first. A successful upload does not prove every page, table, or diagram was read correctly.

Optical character recognition extracts text from an image, such as a scan or photograph of a page.

Everyday example: A photo of a printed recipe becomes text you can select or search.

Keep in mind: Compare names, numbers, and punctuation with the image. A misread character can change the meaning.

Using a microphone to provide input instead of typing. Some tools transcribe speech; others support a spoken conversation.

Everyday example: You say, “Help me organize these three tasks,” then review the response.

Keep in mind: Check that names, dates, and numbers were heard correctly. The service’s recording and privacy controls vary.

An AI system that can select steps and use available tools to pursue a goal. Products use the word “agent” in different ways.

Everyday example: An agent might look up information, organize findings, and prepare a draft.

Keep in mind: Find out whether it can only draft or can also send, buy, delete, or change settings. Set clear review points.

A process arranged to run after a trigger or schedule, with or without AI.

Everyday example: A weekly reminder appears at a chosen time.

Keep in mind: Confirm the trigger, destination, and stopping method. A suggested schedule in chat is not proof an automation was created.

The rights you give an app to read information or perform actions.

Everyday example: A tool asks to view a calendar or edit files.

Keep in mind: Read the scope and use only the access the task needs. Permission to read is different from permission to change or send.