AI for students: which AI is free and best for your studies?

As a student, you are probably the heaviest user of AI and at the same time the one with the smallest budget. The good news for 2026 is that those two barely clash anymore. The free versions from the major providers have improved so much this year that paying is no longer necessary for most study tasks, and where it does pay off, there are student rates of around five euros.

This article covers three things: what you get for free and where those free tiers stop, which student discounts exist and who they are worth it for, and where the line lies between AI as a study aid and AI as a ghostwriter. That last part is less fun but more important than the rest.

What do you already get for free?

The free tiers became a serious product in 2026, and that is a break from last year. Three things have changed.

Since 6 August 2026, free ChatGPT users can chat without limits on GPT-5.6 Luna, with a Think button for heavier questions. There are still limits on uploads, images and tools, but for pure text you no longer hit a wall. For comparison: previously, a free user got roughly ten messages per five hours on the top model.

Claude uses Claude Sonnet 5 in its free tier. That is not a throttled version, but the same model paying users rely on as their daily workhorse, with a context window of one million tokens. For writing and reviewing Dutch texts, it is the strongest free option available.

Gemini, Le Chat and Perplexity also have free versions. With Gemini, you get access to Google's models without a subscription, Le Chat lets you work with Mistral for free, and Perplexity Standard gives you searches with source citations. Exactly what the free tiers can and cannot do is covered in free AI chat in Dutch.

The brake is almost never on the text itself, but on everything around it: the number of files you can upload per day, the use of image generation, and how often you can call the heavy reasoning model. For a student who mostly types and reads, that is rarely a problem.

Which two student discounts are worth it?

Two providers have a real student rate, and both are around five euros per month.

Perplexity Education Pro is the more interesting of the two if you do research. The difference from a regular chatbot is that every answer comes with sources, so you can click straight through to the original article instead of having to trust a summary. For a literature review, that saves hours, and more importantly, it makes what you write verifiable. Perplexity AI: the AI that does cite its sources explains where that approach is strong and where it is not.

Le Chat Student is attractive for a different reason. Mistral processes data in the EU, in France and Sweden, and the Pro subscription unlocks a mode without telemetry. If you work with research data, interview transcripts or personal data, that is not a luxury but a requirement. The context window of Mistral Medium 3.5 is smaller, though, at 256K tokens compared to one million with Claude and Gemini, so it is not the best choice for analyzing entire books.

What does not exist: a student discount at OpenAI, Anthropic or Google. If you want more than the free tier there, you pay the full rate. Google AI Plus, at €7.99 per month, is still the friendliest entry price, and you get 200 GB of storage with it, which is no small thing for a study account full of PDFs.

Which AI should you use for which study task?

Models are not interchangeable. This is what works best in practice for each task.

Understanding the material. Ask a model to explain a concept at three levels: as you would to a child, as in your textbook, and as on your exam. Any free chatbot can do this. Model choice matters least here.

Finding and checking literature. Perplexity, for the source citations. Important: always check whether the source actually says what the model claims. A link is not proof that the summary is correct.

Writing and reviewing Dutch texts. Claude. In our language test, Claude scores 5 out of 5 on natural Dutch, compared to a 4 for ChatGPT, Gemini, Perplexity and Le Chat. The difference lies in idiom and tone: where ChatGPT sometimes translates an English construction literally, Claude chooses the Dutch phrasing. For a report a lecturer will read, that is the difference between a text that sounds like your own and a text that sounds translated.

Working through large documents. Google AI Pro (€21.99) with NotebookLM and a context window of one million tokens is strongest here, but that is a hefty amount for a student. Free Claude also has one million tokens of context, so try that first.

Programming assignments. Claude Pro includes Claude Code, and GitHub Copilot Pro costs $10 (€10.51) per month. Which approach works when is covered in coding with AI.

Learning languages and translating. Gemini is strong here: accurate and good at translating, although the tone is flat. For a practice conversation, that does not matter.

Study aid or ghostwriter: where is the line?

This is the part you should not skip.

AI is excellent at speeding up thinking you are already doing yourself: explaining what you do not understand, pushing back on your structure, pointing out gaps in your argument, checking your text for spelling and repetition, and forcing you to state your own thesis more sharply. In all those cases, the thinking stays yours and you use the model as a critical fellow student who always has time.

It goes wrong the moment you flip the question: not "help me make this better", but "write this for me". Then you hand in work you did not make yourself and whose reasoning you cannot defend. Here are three practical reasons why that is a bad idea, apart from the ethics.

First, programs have rules about this, and they differ per institution and sometimes per course. Some lecturers explicitly ask you to account for your AI use, others ban it for specific assignments. It gets checked, and the burden of proof that you did the work yourself lies with you. Read the regulations before you start, not after you have been invited for a meeting.

Second, language models are confident when they are wrong. They invent sources, dates and quotes that look plausible. If you have not thought the text through yourself, you will not notice, and it ends up in the work you hand in.

Third, and in the long run the most expensive: you learn nothing from it. An exam, an oral exam or a thesis defense tests what you know, not what the model knew.

Should you pay for AI as a student at all?

Do the math. A regular subscription costs €21.02 per month, or more than €250 per year. If you stack ChatGPT Plus, Claude Pro and Google AI Pro, you are at €64.03 per month and €768.36 per year. For most students, that is hard to justify next to tuition and rent.

The honest advice comes in three steps. Start with the free tiers and use them seriously for a month. If you do not run into anything, that is your answer. If you do run into something, look at exactly where: if it is source research, get Perplexity Education Pro for €5.25. If it is privacy or EU processing, get Le Chat Student for €6.30. If it is storage and document work, Google AI Plus for €7.99 with 200 GB included is the logical step.

Only when you notice that you are structurally switching back and forth between three or four models does a bundled subscription become interesting. That is exactly why we exist: one Dutch subscription with all the models included, invoices in euros, fixed limits. For a student who gets by on the free tiers, it is unnecessary, and we would rather say that honestly than have you take out a subscription you do not use.

If you want to know exactly what each plan costs, see what an AI subscription costs or the complete comparison. If you mainly use AI on your phone between lectures, the best AI chat app for iPhone and Android is more relevant.

Conclusion