AI, explained in plain English

What is generative AI (LLMs explained)

Generative AI is AI that creates new things: text, images, audio or video. ChatGPT, Claude and Gemini are the best-known chat apps for text, and the program inside them that actually writes the sentences is called an LLM (large language model). An LLM has read a huge amount of writing and learned which words tend to follow which. When you ask a question, it builds an answer one word at a time from those patterns. That makes it strong at writing, summarizing and translating, and weaker at getting exact facts right.

Getting started with AI (sign-up to first prompt)

ChatGPT, Claude and Gemini all have a website and a phone app, and you can sign up and start for free. After signing in, type your question in the box at the bottom and send it. To add a photo or document, use the add button next to the box (usually a plus sign). Menu names and free limits change often, so this guide sticks to the steps the three services share.

How to write good prompts (examples)

A prompt is the instruction you type to an AI. To write a good one, say who will read the result, what the situation is, what you want, and what format and length you want back. Anthropic, OpenAI and Google all recommend the same basics in their guides: clear instructions, background, examples and a stated format. Do not try to get it perfect in one go. Read the answer and ask for changes.

Hallucination (why AI makes things up)

An AI hallucination is when an AI states something false as if it were fact. It can invent a book title, a court case that never happened, or a statistic, and attach a believable source. This happens because a chatbot is not a lookup tool. It is a program that writes the most plausible next words. Newer models do it less often, but none have stopped, so check every number and source yourself.

Secret commands (/human, /x10think: real?)

Most "secret commands" that go viral, like /human, /x10think, /ELI5 and /TLDR, are not built-in. The AI simply reads them as words and guesses what you mean. They can work, but for the same reason that "explain it simply" works: it is a plain request, not a command. Real built-in commands, like /compact and /clear in Claude Code, are handled by the program itself. If you need a reliable result, a full sentence beats a shortcut.

Tokens (word count, limits, pricing)

A token is the smallest piece of text an AI works with. It can be a whole word, part of a word, a single character, or a symbol. The AI's reading limit (the context window), its writing limit, and API pricing are all counted in tokens, not words or characters. The same sentence can take a different number of tokens depending on the language, and non-English text often takes more than English.

AI agents (vs. chatbots)

An AI agent is an AI that takes a goal, decides the steps itself, and uses tools to carry them out. A chatbot gives one answer to one question. An agent can take a request like "research three competitors, put them in a table, and make a file" and work through searching, reading, organizing and file creation in order. That is convenient, but mistakes and costs grow with it, so keep payments and accounts out of its hands.

MCP (connecting AI to other apps)

MCP (Model Context Protocol) is a shared standard for connecting AI apps to other services such as Google Drive, Notion and your calendar. The official documentation compares it to a USB-C port for AI: instead of a different plug for every device, one kind of plug works everywhere. You may already be using it without knowing the name, under a menu called "connectors" or "integrations."

Vibe coding (building without coding)

Vibe coding is building software by telling an AI what you want in plain language, looking at what it produces, and fixing it by describing changes, without writing or reading the code yourself. What matters is only whether the result looks and works the way you want. Small personal pages and games are doable without any coding background. Anything that handles money or other people's personal data has clear limits.

AI glossary (30 terms)

Most AI terms in the news fit into four groups: what AI is (basics), what runs inside it (models), how people use it (usage), and what can go wrong (safety). The 30 terms below follow that order. Where we have a longer article on a term, it is linked at the end.

ChatGPT vs Claude vs Gemini

For everyday questions, writing and summaries, any of the three will do about equally well. The differences are in the features around the chat. If you work in Google apps, look at Gemini. If you need to generate pictures, ChatGPT and Gemini both offer image generation, while Claude does not make photos or illustrations. If you read long documents and edit text a lot, many people try Claude first. The most reliable way to choose is to paste the same prompt into all three and compare.

Free vs paid AI: do you need to subscribe?

Most people should start with the free plan. Asking questions, polishing writing, summarizing and quick translation all work on free tiers. Paid plans start to make sense when you hit usage limits often, work with long documents, generate lots of images or video, or use AI every day for work. If you subscribe, try one month and cancel if you don't use it. Cancel in the same place you paid: the website, the App Store or Google Play.

ChatGPT Plus vs Pro

If you ask a handful of questions a day and polish some writing or documents, Plus is enough. Pro is for people who hit Plus limits every week, or who run long documents and research all day. If free feels cramped but Plus feels like too much, there is a cheaper step called Go. Prices change often and I couldn't open OpenAI's pricing page to confirm them, so check the upgrade screen in the app before you pay.

Claude Code vs Cursor

If you want to keep the code open, watch it change and accept edits one at a time, pick Cursor. If you would rather hand over a whole job, like "build this feature and run the tests," and only check the result, pick Claude Code. Claude Code started in the terminal. It now also runs from VS Code and JetBrains extensions, a desktop app and the web, all on the same engine. Cursor is built around calling an agent from inside an editor. If you barely code and want to make one web page, favor whichever shows the result on screen right away. That is usually less frustrating.

Claude Pro vs Max

If you mostly chat and work on documents, Pro is enough. If you run Claude Code all day, or you hit Pro limits every week, Max is the next step. Max starts at 5 times the usage of Pro and includes $100 a month in API credits. Prices below come from claude.com/pricing as of November 5, 2026, and they change, so look at the page once more before paying.

Codex vs Claude Code

If you already pay for a ChatGPT plan and do not want another subscription, use Codex. If you have a Claude subscription, or you prefer how Claude models answer, use Claude Code. By the official docs, the two tools show up in almost the same places (terminal, editor extension, desktop, web or cloud), and both have rules files, MCP, skills, hooks and subagents. A feature list will not settle it. What settles it is which company's account you already pay for, and which model suits your work.

Copilot vs ChatGPT

If you want AI inside Word, Excel, PowerPoint and Outlook, use Copilot. For general conversation, such as asking questions, drafting and analyzing material, ChatGPT is the broader tool. The confusing part is that there are three Copilots: the free one you get by signing in, the one bundled with a personal Microsoft 365 subscription, and Microsoft 365 Copilot, which your employer buys. They are not the same product.

Nano Banana vs ChatGPT images

If you often upload your own photos to swap a background, combine several pictures or edit while keeping the same person, try Nano Banana in the Gemini app first. If you often want images with text in them, or like to keep talking and refining drafts, try ChatGPT first. Which one is better flips whenever a model changes, so this article sticks to facts that stay put (names, how limits work, how watermarks work) and tells you to run the same request through both for quality.

Gemini Notebook (NotebookLM) vs ChatGPT

For studying from your own material, such as textbooks, papers and lecture notes, use Gemini Notebook, which you may still know as NotebookLM. For questions beyond your sources, writing from scratch, brainstorming or code, use ChatGPT. Google renamed NotebookLM to Gemini Notebook in July 2026. It's the same product, and your existing notebooks and sources carry over.

Perplexity vs ChatGPT search

If you want one question answered with several source links that you can click and check, use Perplexity. If you search and then want to keep going in the same conversation, writing, building a table or translating, use ChatGPT. Perplexity opens on a search answer. ChatGPT opens on a conversation with search attached. Either way, a link does not make the content correct, so open the link and check numbers and dates yourself.

Seedance vs Veo vs Kling

For a long single shot (up to 30 seconds), use Seedance 2.5 on Higgsfield. If you have a Google account and want to make a video in the app with no extra signup, use the video feature in the Gemini app. If you want to try the newest version first, use Kling. In the Gemini app, Veo has been replaced by Gemini Omni, so I could not confirm any way to choose Veo inside the app. OpenAI announced Sora's shutdown in March 2026, so it is left out.

AI privacy: how to turn off training

Never paste Social Security or passport numbers, card or bank details, passwords, confidential company files, or other people's health or customer data into an AI chatbot, whatever your settings are. Turning off training doesn't stop chats from being stored for a while. You can switch off training use of your chats in each service's settings, and temporary or incognito chats reduce what is kept. For work, use an account your employer approves, not a personal one.

Deepfake and voice-clone scams

Scammers use AI to copy a family member's voice or a friend's face, and the FTC and FBI have both warned about it. Don't try to tell real from fake by ear or eye. If a call asks for money, gift cards, crypto or personal information, hang up and call the person back on a number you already have. Agree on a family safe word ahead of time. If you've sent money, contact your bank right away and report it.

AI video: monetization and labels

Using AI does not by itself block YouTube monetization. YouTube's help page says that disclosing AI content alone will not limit a video's audience or affect its eligibility to earn money. What can block it is mass-produced, repetitive content, which falls under YouTube's "inauthentic content" policy. Separately, YouTube, Instagram and TikTok all ask you to label AI video that looks realistic.

Make AI images (start free)

Open a chat in ChatGPT or Gemini and describe what you want: the subject, the scene, and the style. The AI draws it. You do not need extra software, only an account. In Google's Gemini help page, image generation with Nano Banana is listed for every tier, including people with no paid plan. Free limits change often, so check the app. The easiest way to start is to make one image, then fix the parts you do not like by asking in plain words.

AI video for beginners: picking a tool

There are two ways to make AI video: write a description and let the AI generate a clip (text-to-video), or start from a photo and make it move (image-to-video). If you are new, image-to-video is easier to predict. Good video generation usually needs a paid plan or credits, and one clip is typically 4 to 15 seconds long. The usual workflow is to make several short clips and join them in an editing app.

AI at work (email, notes, spreadsheets)

AI helps most at work with email drafts, meeting notes, summaries of long documents, translation, spreadsheet formulas and data cleanup, and slide outlines. The method is the same each time: paste in the material and say who reads the result, what it is for, and what format you want. Before you paste anything from your company, check your company's AI policy. Treat the result as a draft, and verify numbers and facts yourself.