AI, explained in plain English

What the different kinds of AI do, which names are models and which are apps, how they link together, and which tools to try. No tech background needed.

⏱ 7-minute read24 kinds of AI toolsUpdated October 2026
Part 1 of 6The basics

The one idea that makes it all click

📱

The app

What you open and use: the website, phone app or button inside your email.

Like a car. ChatGPT, Canva and Otter are apps.

→runs on
⚙️

The model (the engine)

The trained AI inside that does the actual work: writing, drawing, listening.

Like the engine. GPT, Claude, Gemini and Veo are models.

From here on we’ll just say model. One app can use several models, and the same model can power many apps. That’s why names get confusing: ChatGPT is the app, GPT is its model. Good news: you can usually just pick the app that suits your task and let it handle the models.

Model, app, or both?

The names you hear fall into three groups. Knowing which is which explains most of the confusion.

Model and app

They make their own model and their own app to use it.

App only

A handy app built on top of someone else’s model.

Model only

You don’t open these directly. You use them through an app.

Part 2 of 6How it all connects

How they link together

Apps and models connect in three ways. You’ll often see all three in one app.

1

Built in

The app quietly rents a model from an AI company and runs it for you. You never see or choose it.

Notion AIruns onGPTClaude

Most apps work this way. When the model’s maker improves it, the app often gets better too.

2

Model picker

The app has a menu so you can choose which company’s model answers you.

PerplexityPoeCursorlets you pickGPTClaudeGemini

This is the way to use Claude and GPT side by side in one app.

3

Connectors (MCP)

Your assistant reaches into other apps and uses them for you, with your permission.

ClaudeChatGPTconnects toRunwayCanvaGmailNotion

MCP is the plug standard behind this. Think of it as USB-C for AI: one shape, so the same connector works in many assistants.

An assistant connected to other apps 🎬 Runway 🎨 Canva 📧 Gmail 🗂️ Notion Your assistantClaude or ChatGPTthinks with its own model connectorconnector
The assistant does the thinking. Each connected app still does its own job with its own model.

Putting it together

Say you run a café and ask Claude for “a 10-second ad with a voice-over.” Here’s what happens behind the scenes:

1

Plan & write

Claude’s own model writes the script and describes each shot.

2

Make the clip

Through the Runway connector, Runway’s models turn those descriptions into video.

3

Add the voice

A voice connector like ElevenLabs reads the script. Without connectors, you’d copy and paste between three apps yourself.

Part 3 of 6

Common questions

The things people most often ask or mix up, from how apps connect to privacy and cost.

Apps, models & how they connect

Why use Runway inside Claude if Runway has its own app?

The video comes from the same Runway models either way, and it uses your Runway account and credits. What changes is who does the steps.

Inside Claude (via the connector)Claude plans the scenes, writes the prompts, sends them to Runway, checks the clips and moves on to the next step, all in one conversation. Best when the video is one part of a bigger job, like “turn this blog post into a 20-second ad with a voice-over.”
In Runway’s own appYou get every button: timelines, brushes, camera controls, side-by-side previews. Best when you want to fine-tune the shot yourself.

Rule of thumb: use the assistant when you want things done for you across several tools. Open the tool’s own app when you want hands-on control.

Can I use Claude inside ChatGPT with MCP?

Not really. MCP plugs tools and data into an assistant (your files, Runway, Gmail, a website). It doesn’t swap the model doing the thinking. ChatGPT always thinks with OpenAI’s GPT models, and Claude always thinks with Claude models.

A developer could build a custom connector that forwards questions to Claude, but that’s a workaround: you pay for both, and Claude’s answer gets relayed through ChatGPT.

What actually works: use an app with a model picker (like Perplexity, Poe or Cursor) to switch models in one place, or keep both apps open. Since MCP is a shared standard, the same connectors, like Runway or Notion, can often be added to both ChatGPT and Claude.

If ChatGPT and Claude can use the same connectors, what’s the difference?

The model doing the thinking. Both can reach Runway or your Google Drive, but each plans, writes and reasons in its own way. Many people use one for everyday work and try the other when a result isn’t right.

Is Microsoft Copilot the same as ChatGPT?

No. They’re different apps from different companies. Microsoft Copilot runs largely on OpenAI’s GPT models alongside Microsoft’s own, so answers can feel similar, but Copilot is built into Windows, Edge, Word, Excel and Outlook. GitHub Copilot is a separate product just for writing code.

Is the Gemini app the same as the AI answers in Google Search?

Both use Google’s Gemini models, but they do different jobs. The AI answers at the top of Google Search summarize web results for one search. The Gemini app is a full assistant you can chat with, give files to and ask follow-ups.

What’s the difference between Claude, Claude Code and the Claude API?

Same family of models, three ways to use it. Claude is the chat app. Claude Code works directly in a coding project. The API is the doorway developers use to put Claude inside their own apps. OpenAI has the same split with ChatGPT, Codex and its API.

Why does the same model give different answers in different apps?

Each app adds its own hidden instructions, tools and settings, and some use a smaller, cheaper version of the model to save costs. Same model, different setup, different results.

What are custom GPTs, Gems and Projects?

Saved setups with your own instructions and files, so you don’t repeat yourself. They run on the same model underneath. They aren’t new models.

People call some tools “wrappers”. Is that bad?

A wrapper is an app built on someone else’s model (an “App only” tool). That’s normal: you’re paying for a better design, a ready-made workflow and features for one job. Just check whether a general chatbot already does the same thing for you.

Using AI day to day

Is “AI” the same thing as ChatGPT?

No. AI is the whole field. ChatGPT is one app that uses one family of models (GPT). Spam filters, map routes and phone cameras all use AI too.

Why do I get a different answer when I ask the same thing twice?

AI is built to vary its wording a little each time, so you can ask again for a fresh take. If the facts change between answers, treat both with caution and check a reliable source.

Why doesn’t it know about recent news?

A model learns from information collected up to a certain date, so it doesn’t know what happened after. Apps with web search turned on can look things up. Watch for source links.

Does it remember me between chats?

Only if the app has a memory feature switched on. Otherwise each new chat starts fresh. Apps with memory let you view, edit or delete what’s saved.

Does uploading a file teach it for good?

Usually not. It reads your file to answer in that conversation. Whether files are stored is a privacy setting, not learning.

Why are free and paid answers different?

Paid plans usually use stronger models, allow longer files and chats, and add features like video or agents. Free versions are often a smaller, faster model.

Can I run AI on my own computer?

Yes, with open models like Qwen, DeepSeek or Mistral and a free app such as LM Studio or Ollama. It keeps everything private, but you need a fairly powerful computer, and they’re usually less capable than the top online models.

Where should a beginner start?

Pick one general chatbot (ChatGPT, Claude or Gemini) and use it for a week on real tasks. Add specialist tools only when you hit a limit.

Trust, privacy & cost

Can I trust what it tells me?

Mostly, but not blindly. AI can “hallucinate”: state something false with total confidence. Check facts, numbers, quotes and anything medical, legal or financial.

Is it safe to paste private information?

Treat it like email to a company. Avoid passwords, ID numbers and other people’s private details, and check the app’s privacy settings.

Are my chats used to train the AI?

It depends on the app and plan. Many consumer apps let you switch this off in settings, and business plans usually don’t train on your data by default. Check before sharing anything sensitive.

Is connecting apps safe?

You choose what to connect and approve the access, just like “Sign in with Google.” A connected assistant can only see and do what that permission allows. Check before it sends, posts, buys or deletes, and disconnect anything you stop using.

Do I own what AI creates for me?

It depends on the tool’s terms and your country. Most tools let you use what you make, but some limit commercial use on free plans, and purely AI-made work may not get full copyright protection everywhere. Read the terms before selling anything.

Do I pay twice when I connect tools?

Often, yes. You need your assistant’s plan plus an account with the connected service, and work done through the connector uses that service’s plan or credits.

Part 4 of 6

AI tools by type

24 kinds of AI tool, each with examples to try. Every example is labelled so you can see what it is and how you use it.

App

Something you open and use. It runs on a model from another company.

Both

The company makes its own model and its own app for it.

Model

The AI on its own. You can’t open it directly; you use it through an app.

🔌

Works inside apps you already use, like Gmail, Word, Zoom, Notion or your browser. No tool marked 🔌? It works on its own.

Tap any tool to see its listing from our directory: what it’s best for, pricing and similar tools.

No match. Try a simpler word, like “photo” or “write”.

Ask, write & learn

Tools you talk to, type into, or learn with.

Create

Make pictures, videos, voices and music from a description.

Listen & understand

AI that takes in sound, pictures or another language and makes sense of it.

Build & automate

Get AI to make software or do repeat work for you.

Automations

“When this happens, do that”: steps that run on their own, with AI handling parts like summarizing.

Use it for: Sorting emails, saving form replies to a sheet, posting updates.

Automations follow the steps you set. Agents decide the steps themselves.

Browse all 14 in Business & marketing

Business, behind the scenes

AI you rarely chat with, but use every day in apps and shops.

Recommendations

The “you might also like” behind shops, streaming and music apps.

Use it for: Suggesting products, shows, songs or articles people are likely to enjoy.

These pick from real items. A chatbot listing suggestions might invent some.

Browse all 14 in Business & marketing
Part 5 of 6

The models behind the apps

You don’t need to choose these yourself, but you’ll see the names everywhere. Here’s who makes what, and where you’ll meet it.

Text & chat

Large language models (LLMs): they read and write.

GPTOpenAIFound in: ChatGPT, Microsoft Copilot
ClaudeAnthropicFound in: Claude
GeminiGoogleFound in: Gemini, Google apps
MuseMetaFound in: Meta AI (older Llama models are free to download)
GrokxAIFound in: Grok, X
DeepSeekDeepSeekFound in: DeepSeek (free to download)
MistralMistral AIFound in: Le Chat
QwenAlibabaFound in: Qwen (free to download)

Pictures

Image models: they draw and edit.

GPT ImageOpenAIFound in: ChatGPT
Imagen & Nano BananaGoogleFound in: Gemini
MidjourneyMidjourneyFound in: Midjourney
FLUXBlack Forest LabsFound in: Many image apps
FireflyAdobeFound in: Photoshop, Firefly
Stable DiffusionStability AIFound in: DreamStudio

Video

Video models: they animate.

VeoGoogleFound in: Gemini, Google Flow
KlingKuaishouFound in: Kling
Gen seriesRunwayFound in: Runway
HailuoMiniMaxFound in: Hailuo
RayLuma AIFound in: Luma Dream Machine

Voice & sound

Speech and audio models: they listen and speak.

ElevenLabs voicesElevenLabsFound in: ElevenLabs
WhisperOpenAIFound in: Many transcription apps (free to download)
SunoSunoFound in: Suno
Big vs. smallBig models are smarter but slower and pricier. Small ones are quick and cheap, and some run right on your phone.
“Thinking” modelsSome take a moment to reason step by step first. Slower, but better at math, planning and tricky questions.
Open vs. closedOpen models (like DeepSeek, Qwen and Mistral) can be downloaded and run privately. Closed ones are used through the maker’s app.
Part 6 of 6

Jargon buster

Prompt
What you type or say to the AI. Clear, specific prompts get better results.
Model
The trained AI that does the actual work inside an app. Also called the engine.
LLM
Large language model: a model that reads and writes text.
Multimodal
Can handle more than one kind of input, like text plus photos or voice.
Hallucination
When AI states something false as if it were true.
Reasoning model
A model that “thinks” step by step before answering. Slower, better at tricky problems.
Open model
A model anyone can download and run on their own computer, like DeepSeek or Qwen.
Agent
AI that takes actions toward a goal, not just answers.
RAG
Looking things up in your documents first, then answering from them.
Connector
A link that lets your assistant use another app, like Runway or Gmail, for you.
MCP
Model Context Protocol: the shared plug standard behind connectors. Works across Claude, ChatGPT and more.
API
The behind-the-scenes doorway apps use to rent an AI model from its maker.
Model picker
A menu inside an app that lets you choose which model answers.
Training
How a model learns, from huge amounts of examples, before you ever use it.

Examples are starting points, not rankings or ads. Tools change quickly, so features and prices vary by plan and country.

Ready to pick a tool?

Browse 303 AI tools sorted by what you want to get done, with prices shown up front.

Browse all tools