AI Glossary: 25 Terms Explained Without Jargon

AI-assisted content — see our editorial standards

Quick answer

AI conversations are full of words that sound technical but aren't — once someone explains them. This glossary covers the 25 terms you'll actually encounter, defined the way a patient friend would explain them over coffee.

The 25 terms

Software that can perform tasks we used to think required human intelligence — understanding language, answering questions, recognizing images, and holding conversations. The AI you'll use day-to-day is text-based: you type a question, it types back an answer.

The specific type of AI that powers ChatGPT, Claude, Gemini, and most modern AI assistants. It's called "large" because it was trained on an enormous amount of text. It's called "language model" because it learned to predict and generate language — which is why it can hold a conversation. You don't need to remember this term; it just explains why these tools are good at language tasks.

Your question, instruction, or request to an AI. When you type "Can you explain what a Medicare EOB is?" into ChatGPT, that's your prompt. A longer, more specific prompt generally gets a better answer. "Explain what a Medicare EOB is in simple terms, as if explaining to someone who's never seen one before" is a better prompt than just "what is an EOB."

When an AI confidently states something that is false. This is one of the most important limitations of current AI. The AI isn't lying — it doesn't "know" it's wrong. It's generating what sounds like a plausible answer based on patterns it learned, and sometimes those patterns lead it somewhere inaccurate.

Always verify specific facts — especially numbers, dates, and legal or medical rules — from authoritative sources. A hallucination can sound exactly as confident as a correct answer.

How much of a conversation the AI can "remember" at once. Each AI model has a limit on how much text it can hold in mind — called its context window. If you have a very long conversation, the AI may "forget" the earlier parts of it. In practical terms: if you're discussing something important from early in a conversation, you may need to remind the AI of it.

The unit AI uses to measure text. Roughly speaking, one token equals about three-quarters of a word. AI tools often limit or charge by token usage — which is why some free tiers have usage limits. You don't need to count tokens yourself, but it's why very long conversations or large documents can hit limits faster.

The no-cost version of an AI service. Most major AI tools — ChatGPT, Claude, Gemini, Copilot — offer a free tier that's genuinely useful. Paid plans typically add faster responses, higher usage limits, or access to newer AI models. See our guide: Is AI free? What ChatGPT actually costs.

Instructions given to the AI before a conversation starts that shape how it behaves. When you use our Document Analyzer, for example, the system prompt tells the AI to focus on Medicare EOBs, insurance documents, and financial statements — before you say anything. You never see the system prompt; it works in the background.

An older term for software that responds to typed questions. Traditional chatbots followed rigid scripts and couldn't handle anything unexpected. Modern AI assistants like ChatGPT are far more capable — they can handle novel questions, nuance, and ambiguity. When people call ChatGPT a "chatbot," they usually mean it in this broader, more impressive sense.

The specific AI brain running behind a tool. OpenAI's GPT-4o, Anthropic's Claude 3.5 Sonnet, and Google's Gemini 2.0 are all different models. Different models have different strengths, speeds, and costs. When an AI company releases an "upgrade," they're releasing a new model. As a user, you generally don't need to choose — the tool picks the best model available for your plan.

The text (and other information) the AI learned from before it was deployed. Modern AI models were trained on enormous amounts of text from the internet, books, and other sources. This is why they know so much — but it's also why their knowledge has a cutoff date and why they can occasionally reflect biases present in that training data.

The date after which an AI has no training data. Events, price changes, law changes, and new research that happened after this date may not be reflected in the AI's answers. ChatGPT and Claude typically display their knowledge cutoff date in their settings or documentation. For current information, use an AI with web search enabled (like Gemini or Perplexity), or verify with official sources.

Some AI tools can search the internet in real time to find current information — current Medicare premiums, recent drug recalls, today's weather. ChatGPT's free version now includes limited web browsing. Gemini searches the web by default. If an AI has this feature enabled, it will usually note that it searched the web when giving you a result.

AI that can work with more than just text — including images, audio, and documents. Most major AI tools in 2026 are multimodal: you can show them a photo and ask what's in it, upload a PDF and ask for a summary, or speak your question instead of typing it. Our Document Analyzer uses this capability — you upload a PDF and the AI reads and explains it.

AI's ability to look at a photo and understand what's in it. This is how your phone recognizes faces, how Google Photos finds photos from your trip to Yellowstone when you search "Yellowstone," and how the Be My Eyes app can read a prescription label aloud to you.

The technical field of helping computers understand human language. NLP is why you can type a question in everyday English and get a useful answer — rather than needing to type in computer-style commands. All modern AI assistants use NLP. You don't need to remember this term; it's background information.

Technology that converts written text into spoken audio. Many AI tools, including ChatGPT's mobile app, can read their responses aloud. This is useful if you find it easier to listen than to read, or if you want your hands free while getting information.

The reverse of text-to-speech: your spoken words are converted to text. The ChatGPT mobile app lets you speak your question rather than type it — your voice is converted to text, sent to the AI, and the AI's answer appears (or can be read aloud to you). This is particularly useful for anyone who finds typing difficult.

AI that creates new content — text, images, audio, video — rather than just analyzing or classifying existing content. ChatGPT, Claude, and Gemini are all generative AI: they generate new text in response to your question rather than just looking up a stored answer. Image generators like Midjourney create images from text descriptions.

A set of rules or instructions a computer follows to accomplish a task. AI systems use algorithms to process information and generate responses. You'll hear this word often in discussions about AI, social media feeds, and search engines. In plain English: it's how the computer decides what to do.

How your information is handled by the companies behind AI tools. Key questions: What gets stored? Who can access it? Is it used to train future AI models? These questions matter most when you're sharing personal or medical information with AI tools. For how our tools handle your data, see our Privacy Policy.

The technology that lets different software talk to each other. When our Document Analyzer sends your PDF to Claude for analysis, it does so through Claude's API. You'll see this word in technical discussions about how AI tools are built, but you never need to interact with an API yourself as a regular user.

A category of AI tool that works alongside specific software — like Microsoft Copilot, which is embedded in Windows, Word, Excel, and Outlook. The word "copilot" is used to suggest the AI is your assistant in a specific context, not a standalone tool.

A catch-all term for AI tools designed to help with everyday tasks — answering questions, writing, planning, explaining. ChatGPT, Claude, Gemini, and Copilot are all AI assistants. The term distinguishes them from more specialized AI (like image generators or medical diagnostic AI).

AI-generated media — video, audio, or images — that appears real but is fabricated. Deepfakes can make it look or sound like someone said or did something they never did. This technology is increasingly used in scams — including voice cloning scams where a caller mimics the voice of a family member. See our guide: What is a deepfake? A guide for people over 50.

Frequently asked questions

No. The most important ones are: prompt (what you type), hallucination (when AI gets something wrong), and free tier (you can start without paying). The rest are useful context but won't affect your day-to-day use.

AI tools default to the vocabulary level of the average person they were trained on, which tends to be somewhat technical. If an AI uses a word or phrase you don't recognize, just ask: "Can you explain what [term] means?" It will clarify in plain English.

No. AI is the broad category; ChatGPT is one specific AI tool made by OpenAI. Saying "AI" and "ChatGPT" interchangeably is like saying "car" and "Toyota Camry" interchangeably. ChatGPT is a type of AI, but AI includes many other tools.

It means the product includes some form of AI capability — though the sophistication varies enormously. A spam filter that blocks junk email "uses AI." A conversational assistant that can answer complex questions and write letters also "uses AI." The term has become so broad as to be nearly meaningless without specifics.

The most important terms to remember

If you only take three things from this glossary:

1. Prompt — what you type to the AI. More specific prompts get better answers.
2. Hallucination — AI can be wrong and sound confident. Verify anything important.
3. Free tier — all the major AI tools are free to start. You don't need to pay to try them.

Everything else is context that helps you understand conversations about AI — but none of it is required to use the tools effectively.

• What is AI, really? A plain-English guide for people over 50
• How to use ChatGPT for the first time
• Is AI safe to use? Privacy, data, and what to know