LLMs for Complete Beginners: A No-Hype Guide

You've probably heard about ChatGPT. Maybe Claude, Gemini, or DeepSeek too. But if you're not sure what these things actually are or why you'd use one over another — this guide is for you.

No hype, no "AI will change everything" rhetoric. Just a practical look at what these tools can do and how to actually use them.

What Are LLMs?

LLM stands for Large Language Model. These are software systems trained on massive amounts of text that can generate human-like responses to your questions and requests.

The main players right now:

ChatGPT (OpenAI) — The one that started the hype. Solid all-rounder.
Claude (Anthropic) — Strong at writing, analysis, and longer documents.
Gemini (Google) — Integrated with Google services, good for research.
DeepSeek (DeepSeek) — Chinese-developed, surprisingly capable, very cheap.
Llama (Meta) — Open source, can run locally on your computer.

They all work similarly: you type a question or instruction, and they generate a response. Think of them as extremely well-read assistants that can write, summarize, explain, and brainstorm.

What they don't do:

Have real-time internet access (unless specifically enabled)
Know anything after their training cutoff date
Have feelings, opinions, or consciousness
Replace human expertise for critical decisions

What LLMs Are Actually Useful For

First Drafts

Staring at a blank page is the worst. Describe what you need, and let the LLM generate a starting point.

"Write a first draft of an email to my landlord about a broken heater. Polite but firm. Mention it's been two weeks."

You'll have something workable in seconds. Edit it to sound like you.

Summarizing Long Stuff

Paste a long article, report, or document and ask for a summary.

"Summarize this in 5 bullet points for a small business owner: [paste text]"

Explaining Complex Topics

Need something explained at your level?

"Explain how HTTPS encryption works like I'm a smart 12-year-old."

Or:

"Explain the EU AI Act to someone who runs a small software company."

Brainstorming

Stuck? LLMs are solid brainstorming partners.

"I run a small bakery. Give me 10 creative Valentine's Day marketing ideas that don't need a big budget."

Proofreading

Paste your text and ask for specific improvements.

"Proofread this email for grammar and suggest how to make it more concise: [paste text]"

Learning

They can explain concepts, provide exercises, and handle follow-up questions.

"I want to learn basic Excel formulas. Give me the 5 most useful ones for managing a small business budget, with examples."

How to Write Better Prompts

The quality of what you get depends heavily on how you ask.

Be specific:

❌ "Write an email"
✅ "Write a 3-paragraph email to a client explaining a one-week project delay. Apologetic but professional."

Give context:

❌ "How do I fix this?"
✅ "I'm getting 'Permission denied' when saving a file in Word on Windows 11. What should I try?"

Assign a role:

❌ "Explain marketing"
✅ "You're a marketing consultant for small businesses. Explain social media basics to a restaurant owner who's never done it."

Request formats:

❌ "Give me tips"
✅ "Give me 5 tips as a numbered list with brief explanations"

Iterate: If the first response isn't right, ask for adjustments: "Make it shorter," "Use simpler language," "Add more detail about X."

Which One Should You Use?

Here's my honest take:

ChatGPT — Best for: general-purpose tasks, coding, creative writing. Free tier gets you GPT-4o mini. The €20/month Plus gives you the full GPT-4o and reasoning models.

Claude — Best for: long documents, analysis, nuanced writing. I find it less likely to give generic responses. Free tier is decent; Pro is €18/month.

Gemini — Best for: research and anything Google-related. It can access your Gmail, Docs, etc. Free tier is solid.

DeepSeek — Best for: budget users who want near-GPT-4 quality at a fraction of the cost. Privacy concerns since it's based in China, so I wouldn't use it for sensitive work.

Local models (Llama, etc.) — Best for: privacy-conscious users willing to set things up. Runs on your hardware, no data leaves your machine.

For most people, I'd start with Claude or ChatGPT's free tier. Try both for a week and see which feels better.

What LLMs Are Bad At

Facts

They can confidently state things that are completely wrong. Always verify:

Statistics and numbers
Recent events
Technical specifications
Legal or medical information

Math

Simple arithmetic is fine. Complex calculations? Use a calculator.

Knowing About You

They don't know your situation unless you tell them in the conversation.

Original Research

They synthesize existing knowledge. They can't run experiments or generate truly novel discoveries.

Privacy Considerations

This matters more than most people realize.

Don't paste sensitive data — Customer info, passwords, proprietary code. Just don't.
Assume conversations are stored — Companies claim not to use your data for training (especially on paid plans), but they can still review conversations.
Consider local alternatives for sensitive work — If privacy really matters, run a model locally (I'll cover this in another post).

Provider privacy comparison:

OpenAI/ChatGPT: Can opt out of training on your data
Anthropic/Claude: Similar policies, generally considered privacy-focused
Google/Gemini: It's Google. Draw your own conclusions.
DeepSeek: Chinese company, subject to Chinese data laws

Free vs. Paid

Free tiers are actually pretty good now:

ChatGPT: GPT-4o mini (capable, some limitations)
Claude: Decent usage limits
Gemini: Good integration with Google services

Paid plans (~€20/month) get you:

Better models with stronger reasoning
Higher usage limits
Advanced features (image generation, file analysis, etc.)

My advice: Start free. If you're using it daily for work and hitting limits, the paid plans are worth it.

Getting Started

Pick one: chat.openai.com, claude.ai, or gemini.google.com
Create a free account
Try a simple prompt: "Help me write a professional out-of-office email for a week-long vacation"
Experiment with follow-ups and refinements

The Bottom Line

LLMs are useful tools for writing, learning, and brainstorming. They're not magic. Treat them like smart assistants that need clear instructions and whose work you should verify.

Used well, they save hours every week. Used carelessly, they waste time or get you in trouble.

Start small. Verify outputs. Expand your use cases as you get comfortable.