How to Use an AI Chatbot: A Beginner's Guide

To use an AI chatbot, you type a question or a request in plain language, the way you’d write to a helpful friend, and it writes back. You get better answers when you give it context (“I’m a nurse writing to a patient’s family”), say what you want back (“three short options”), and keep the conversation going with follow-ups. You don’t need special commands or any technical knowledge.

What you do need to know is where your words go, what these tools are bad at, and how to tell a good answer from a confident wrong one. That’s what this guide covers.

What is an AI chatbot, in plain terms? #

An AI chatbot is a program built on a large language model (LLM). The model was trained on a huge amount of text and learned to predict which words should come next. When you ask it something, it generates a reply one word-piece at a time, based on the patterns it learned.

That explains both its strengths and its failures:

• It’s very good with language: rewriting, summarizing, explaining, brainstorming, drafting emails, translating, answering “how does X work” questions.
• It doesn’t “look things up” unless the app gives it web search. A plain model answers from memory, and that memory stops at a training cutoff date.
• It can sound certain while being wrong. The model writes what sounds like a plausible answer, and plausible isn’t always true. We explain why in why AI makes things up.

Which kind of AI chatbot should you start with? #

There are two kinds, and the difference matters more than which brand you pick.

If you want the smartest answers and don’t mind your chats being stored on a company’s servers, a cloud chatbot is the easy start. If you’d rather keep what you type to yourself, or you want AI on a plane or somewhere with no signal, a local app is the better fit. Plenty of people use both: cloud for general questions, local for anything personal. Our plain-English guide to local LLMs explains how the on-device kind works.

How to start using an AI chatbot, step by step #

Here’s the whole process with a local app, since it has one extra step (downloading a model). With a cloud chatbot you skip step 2 and sign in instead.

1. Install the app. Personal LLM is free on iPhone (iOS 15.1 or later) and Android (7 or later) and needs no account or sign-up.
2. Download a model once, on Wi-Fi. The model is the “brain” file. Personal LLM shows a “Fits your device” badge on each model, read from your phone’s RAM, so you can’t pick one that’s too big by mistake. Qwen 3.5 4B (2.74 GB) is the app’s recommended first download; older phones can start with Gemma 4 E2B (2.04 GB).
3. Open a new chat and type. Ask something you actually want to know. “What should I cook with chicken thighs, rice and a lemon?” is a better first test than “Hello.”
4. Read the answer, then follow up. “Make it spicier.” “I don’t have an oven.” “Shorter, please.” The chatbot remembers the conversation so far, so you don’t need to repeat yourself.
5. Try airplane mode. With a local app, the model keeps answering with no connection at all. That’s the easiest way to see that nothing is leaving your phone.

How do you ask an AI chatbot a good question? #

Most disappointing answers come from short, context-free prompts. The model fills gaps with generic guesses. Give it the details a person would need.

Weak: “Write an email to my landlord.”

Better: “Write a short, polite email to my landlord. The kitchen tap has leaked for two weeks and I reported it on March 3. I’d like a repair date by Friday. Keep it under 120 words and don’t sound angry.”

A few habits make a big difference:

• Say who you are and who it’s for. “Explain this to a 10-year-old” and “explain this to a pharmacist” produce very different answers.
• Say what format you want. A table, a numbered list, three options, one paragraph, a subject line.
• Paste the material. If you want feedback on your cover letter, paste the letter. Don’t describe it.
• Give an example. “Here’s a message I liked; write mine in the same tone.”
• Ask it to ask you. “Before you answer, ask me any questions you need.” This works well for plans and advice.

For a longer list of starting points, our system prompt examples show how to set up a chat for a specific job, such as a recipe helper or a proofreader, so you don’t repeat instructions every time.

What should you check before trusting an answer? #

Treat a chatbot like a well-read friend who never says “I don’t know.” Useful, fast, and sometimes confidently wrong.

• Facts, dates, numbers and quotes: check anything you’ll rely on against an official source.
• Math: small models especially can slip on arithmetic. Ask it to show the steps, or turn on a “thinking” mode if your app has one.
• Recent events: a model without web search doesn’t know what happened after its training cutoff and may guess.
• Links and citations: chatbots sometimes invent sources that look real.
• Medical, legal and money questions: use AI to understand terms and prepare questions, then confirm with a professional.

A good rule: the more it matters, the more you verify.

What shouldn’t you type into an AI chatbot? #

With a cloud chatbot, assume anything you type is stored on the company’s servers and may be reviewed or used for training, depending on your settings. That doesn’t make it dangerous for everyday questions, but it’s a reason to leave out:

• passwords, card numbers and bank details
• Social Security, passport or ID numbers
• medical records or other people’s health details
• confidential work documents, client names or code your employer hasn’t approved
• anything you’d be upset to see in a data breach

Our guide to keeping personal data out of AI chatbots shows how to strip names and numbers from a prompt and which settings to change.

A local app changes that calculation. In Personal LLM the model runs on the phone, there’s no server, and chats live in the app until you delete them. The network is used only to download models from Hugging Face and to show ads in the free version, and ads can’t load when you’re offline. So a local app is a reasonable place to ask about a diagnosis you’re worried about, a draft of a difficult message, or a document you’d never upload. It’s still smart to lock your phone with a passcode, since anyone who can open it can read your chats.

Beginner mistakes to avoid #

• Stopping after one answer. The follow-up is where chatbots shine. Ask for changes until it’s right.
• Asking yes/no questions. “Is this email okay?” gets “Yes, it’s great!” “What would make this email clearer? List three changes” gets useful feedback.
• Believing it knows you. Unless you tell it, the chatbot doesn’t know your age, location, budget or situation.
• Using one giant chat for everything. Start a new chat for each topic. Long chats eventually push early details out of the model’s working memory.
• Forgetting it can refuse or hedge. If an answer is vague, ask it to be specific, or rephrase the question.

Frequently asked questions #

Is using an AI chatbot free? #

Most have a free option. ChatGPT, Gemini and Claude offer free tiers with limits and sell paid plans with better models. Local apps that run open models on your phone are often free; Personal LLM is free with ads and has a one-time purchase to remove them.

Do I need an account to use an AI chatbot? #

Cloud chatbots usually want an account for full features and to save your history. Local apps generally don’t need one, because nothing is stored on a server. That’s one reason people who care about privacy start there.

Can an AI chatbot work without internet? #

Cloud chatbots can’t, because the AI runs on the company’s servers. A local AI app can: once the model file is downloaded, it answers in airplane mode, on a plane or in a basement with no signal.

Is it safe to use AI chatbots? #

For everyday questions, yes, as long as you verify important facts and keep sensitive details out of cloud services. The main risks are wrong answers delivered confidently and personal data stored on someone else’s servers. Checking facts handles the first; a local app or careful prompting handles the second.

What’s the best first thing to ask an AI chatbot? #

Something real from your day: rewrite a message you’re unsure about, plan a week of dinners from what’s in your fridge, or explain a term from a letter you received. Real tasks show you quickly what the tool is good at, far better than trick questions do.