You want to get started, but you don't know where. ChatGPT, Claude, Gemini, Copilot, Le Chat: everyone has an opinion, every article publishes its own ranking, and last week's ranking is already contradicted by this week's.
This guide won't crown a winner. Not out of caution, but because these tools change all the time: new models, new features, free limits raised or lowered. A fixed ranking ages in a few weeks. What doesn't age is a way to choose based on how you use it, and a test you can rerun whenever you like.
The five assistants, in a nutshell
Here's who's who, without details that will have changed by next month:
- ChatGPT, by OpenAI: a general-purpose assistant, often the first one people hear about.
- Claude, by Anthropic: a general-purpose assistant from another American company.
- Gemini, by Google: Google's assistant, designed to fit in with the company's services.
- Microsoft Copilot: Microsoft's assistant, available in several of its products.
- Le Chat, by Mistral AI: the assistant from the French company Mistral AI.
They're all built on an LLM, a large language model, and they all work the same way from your side: you write a request, it answers, you follow up. They can all be confidently wrong, too. If you're mainly torn between the first two, our Claude or ChatGPT comparison digs into that matchup.
What you won't find here (and why)
This article has no prices, no number of free messages per day, no model names and no benchmark scores. That information changes often, sometimes from one week to the next, and an article that freezes it ends up misleading people.
The right habit: check the free plan's terms on each tool's official page at the moment you choose. Be wary of comparisons quoting precise numbers without a date, and of lookalike apps: go through the publisher's official website or app.
Choose by use, not by reputation
The question isn't "what's the best AI?" but "which AI works best for what I actually do?" Start by writing down your three main uses. Here are the most common ones, and what to look at for each.
Writing and rewording (emails, messages, texts): does the tone sound like you? Is the text natural or full of empty phrases? Does the assistant respect a length instruction?
Understanding and learning (explaining a concept, studying): is the explanation clear? Does it admit when it doesn't know? Does it ask you questions when your request is vague?
Finding recent information: some assistants can browse the web, others can't, or only on certain plans. If that's your main use, check that the free version allows it, and see whether it cites sources you can open.
Working with documents (summarizing a PDF, analyzing a spreadsheet): does the free version accept files? Is the summary faithful to the document?
Fitting in with your tools: if you already spend your day in one company's services, its assistant may be more convenient day to day. Still, check what's included for free.
Privacy: each service has its own settings (history, whether your chats are used for training). Look at them before you choose, and read ChatGPT and personal data to know what to look for. The habits apply to every assistant.
Before comparing assistants, also make sure your task really is a text job. An illustration, a piece of code or a task you repeat every week often belongs to other families of tools, and some checks are better done yourself, at the source. To practice that sorting, play Which AI for the job?, a free mini-game that takes a few minutes.
The home test
The only ranking that matters is yours. Here's how to build it.
1. Pick three real tasks. Not riddles to trip up the machine, but things you actually do: an email to write, a concept to understand, a text to summarize.
2. Add a trick question. A factual question you know the exact answer to, on a topic you know well: your town, your job, a hobby. It's the best way to see how an assistant behaves when it doesn't know.
3. Use identical prompts. Same text, word for word, in each assistant. Open a new conversation each time so nothing skews the result.
4. Score with a simple grid. For each answer, a score from 1 to 3 on four criteria:
| Criterion | Question to ask yourself |
|---|---|
| Usefulness | Can I use it as is, or nearly? |
| Accuracy | Are the facts right? Does it flag its doubts? |
| Clarity | Is it easy to read, at the right length? |
| Comfort | Did the interface help me or get in the way? |
5. Follow up once. Ask for a specific change ("shorter," "warmer"). A good assistant for you is also one that understands your tweaks.
Add it up. Your test's winner is a good starting point, not a lifelong commitment: nothing stops you from using two depending on the task.
Mistakes to avoid
- Choosing after a single try. One brilliant answer may be luck; so may one bad answer.
- Comparing different prompts. If you change the question, you're comparing your wording, not the tool.
- Believing an assistant is never wrong. None of them checks its facts on its own. To learn how to check an answer, read how to fact-check an AI answer.
- Switching tools instead of improving your request. Often the problem isn't the assistant but missing context. A precise prompt improves every assistant's answer.
- Trusting an undated ranking. What was true six months ago may not be anymore.
Mini exercise: your own comparison in 15 minutes
- Prepare (3 min): write two prompts in a note. One for a real task from your week, with context. One factual question you know the answer to.
- Test (8 min): paste them, unchanged, into two or three assistants of your choice, each in a new conversation.
- Score (3 min): fill in the grid for each answer.
- Decide (1 min): which one you'll keep this week, and for what.
What to notice: are the gaps big or tiny on your real task? Who got the factual question wrong, and who flagged it? You'll often find that the difference comes more from your request than from the tool. Rerun the test in a few months: the results may have changed.
Going further
Once you've picked your assistant, the guide on how to use ChatGPT for beginners walks you through your first conversation, and it works for any assistant. For a fair test, write good requests: practice in the prompt lab. And to train yourself to spot errors, try Hallucination Hunt.
Choosing your assistant is part of the free chapter "Light the Lantern": you learn what an LLM is, choose your assistant and set it up. It runs in your browser, no account needed: play the chapter.
Frequently asked questions
What's the best free AI assistant?
There's no universal answer, and the tools change very fast. The best one for you is the one that handles your real tasks best: test two or three assistants with exactly the same prompt and score the answers with a simple grid.
Are free versions enough to get started?
Generally yes, for exploring and everyday use. The included features and usage limits vary by tool and change often: check them on each assistant's official page when you choose.
Can I use several AI assistants at once?
Yes. Nothing stops you from keeping one for writing and another for a different task. Just be careful with your data: each service has its own privacy settings, to check one by one.
Do I need an account to use a free AI assistant?
It depends on the tool and the time: some let you start without an account, others require one. An account often gives you history and settings. Terms change regularly, so check them on the official website.