Free ChatGPT in English — Your online AI assistant

Looking for a free ChatGPT in English? Chat Français is a free AI chat you can use right away, with no sign-up and no credit card to get started. Type your question in the box above and the assistant starts answering within seconds: chat, write, proofread and translate, the way you would with ChatGPT, entirely in English.

Here’s the straight answer first: this isn’t ChatGPT, and it isn’t made by OpenAI. Chat Français is an independent AI chatbot, run from France and built on recent large language models. The free version handles everyday questions, writing and summaries; Premium adds more when you need it.

Ask about anything: general knowledge, advice for work, help with homework, a text or an email to write. Whether you’re a student, a professional or just curious, the assistant adapts to your level and to what you need, and there’s nothing to download.

What you can do with a free ChatGPT in English

Mostly small, everyday jobs that would otherwise eat half an hour: a cover letter tailored to a specific job posting, a polite but firm email to your landlord about a repair that keeps getting pushed back, a toast for a friend’s wedding, or a clear explanation of compound interest the night before a finance quiz.

It’s good at practical planning too: a week of budget dinners built around what’s already in your fridge, a three-day itinerary for Chicago, a packing list for a camping trip. For work or school, it can outline a presentation, turn messy meeting notes into a clean summary, explain an XLOOKUP formula, or work out why a few lines of Python throw an error.

Like any AI, it can make mistakes. For health, legal or money decisions, use it to understand the topic and prepare your questions, then check with a professional or an official source.

• writing: cover letters, emails, speeches, posts
• planning: meals, trips, packing lists, budgets
• work and school: outlines, summaries, spreadsheet formulas, code
• health, legal or money questions: understand first, then check with a professional

How to get good answers

Write the way you’d ask a knowledgeable friend, in plain English. Context makes the difference: say who the text is for, how long it should be and what tone you want. If the first answer isn’t quite right, just say so (“shorter”, “more formal”, “give me three options”) and the assistant rewrites it.

When an answer is almost right, follow up on one specific point instead of starting over. “Take point 2 and expand it with an example” works better than asking the whole question again, because the conversation keeps the context you’ve already set.

You can also ask for a format: a bulleted list, a table, a short paragraph, a step-by-step plan. The same content, laid out the way you’ll actually use it, saves you a round of copying and reformatting.

• say who the text is for, how long it should be and what tone you want
• ask for a format: a list, a table, three options, a short paragraph
• follow up on one point instead of rewriting the whole request
• it remembers the conversation, so there’s no need to repeat yourself

An AI that writes natural English, in any register

The assistant answers in English by default, in a natural register rather than a stiff, translated one. Tell it which conventions you need (American or British spelling, dates written month-first or day-first, a formal or a casual tone) and it sticks to them for the rest of the conversation.

Switch to Spanish, German or French mid-conversation and it follows you into that language. That makes it handy for translating a message on the spot, checking how a sentence sounds to a native speaker, or practicing a language you’re learning: ask it to correct your mistakes and explain each one briefly.

Chat Français started as a French service, which is where the name comes from. The chat itself works in English from the first message to the last; a few pages, such as the legal notice, the privacy policy and the sign-in pages, are still in French for now.

• answers in English by default, American or British spelling on request
• follows you if you switch languages mid-conversation
• quick translations that keep the tone: formal, friendly or neutral
• a patient partner for practicing a language you’re learning

On your phone, with nothing to install

Everything runs in your phone’s browser: no app to download, no storage to free up, no updates to manage. That matters more than it sounds, because most real-life uses happen on the go: making sense of a letter you just received, drafting a message, checking a sentence before you send it.

No app also means no permissions up front. Nothing asks for your contacts, your photos or your location, because nothing is installed on the device. The page does what it shows, and nothing more.

For quick access, add the page to your home screen from your browser’s share menu. You get an icon that opens like an app, without being one: nothing to install, no account, no updates.

• works in any mobile browser, on iPhone or Android
• no permission requests: no contacts, no photos, no location
• add the page to your home screen for one-tap access
• the layout adapts to phones, tablets and computers

Send it a photo, a PDF or a spreadsheet

Attach a file of up to 50 MB (a photo or screenshot, a PDF, a Word document, a PowerPoint deck, an Excel sheet or a CSV) and ask what you need: “summarize this lease in plain English and flag anything I should push back on”, “list the key dates in this contract”, “which products sold best in this sheet?”

The assistant reads the file and answers right in the chat. A question about a file counts as a regular message, so the free limits below apply.

When only the content or the wording matters, remove what identifies people first (names, file numbers, contact details). The analysis will be just as good, and you won’t have shared anything you didn’t need to.

• photos, screenshots, PDF, Word, PowerPoint, Excel and CSV, up to 50 MB
• ask for a summary, the key dates or the clauses to watch
• a question about a file counts as a regular message
• remove names and numbers when only the content matters

How many free messages per day, exactly

Without an account, and without giving an email address, you can send 20 messages every 24 hours, within an overall cap of 50 messages over 90 days. No credit card, and no trial that quietly turns into a subscription. A question and its follow-ups fit within a day; regular daily use will reach the limit, and that’s who Premium is for.

The limit renews gradually over a rolling 24 hours, not at midnight: once you reach it, you can write again as the hours go by. If you connect from a shared place (a school, an office, a library), a safety cap shared by every device on the same network also applies, on top of your own limit.

Why a limit at all? Every answer has a real cost, and without a cap a few automated scripts would use up what keeps the service free for everyone. When you reach it, the chat asks for your email address and shows the Premium offer. You don’t have to take it: you can simply come back later.

Some features are built mainly for subscribers. On the free version, creating images and ready-to-download Word, Excel, PowerPoint or PDF files is limited to a small trial, sometimes just a blurred preview. Live web search, for news, scores or prices that change every day, is reserved for Premium; without it, the assistant answers from what it already knows. The chat doesn’t make videos.

• 20 messages per 24 hours without an account, hundreds a day with Premium
• an overall cap of 50 messages over 90 days
• gradual renewal over a rolling 24 hours, not at midnight
• on a shared network, a safety cap is shared by all devices
• images, downloadable files and live web search: mainly Premium

Premium, if you need more

Premium costs €9.99 a month. It brings hundreds of messages a day, image generation, downloadable Word, Excel, PowerPoint and PDF files, live web search, and answers from a more advanced GPT model by OpenAI, up to a very high daily volume.

There is no commitment on the monthly plan: you can cancel at any time and keep Premium until the end of the month already paid. Prices are in euros, so your bank converts the amount if you pay in another currency. And if the free version is enough for you, keep using it: nothing forces you to upgrade.

• €9.99 a month, billed in euros; no commitment, cancel at any time
• hundreds of messages a day, with a more advanced GPT model up to a very high daily volume
• images, downloadable Office and PDF files, live web search

See what Premium includes

Students, professionals, everyday life: where to start

For school or college, the best way in isn’t asking for the homework but for an explanation. Have a lesson you didn’t get explained again, check your understanding before a test, practice on exercises generated on demand: that’s what makes you progress, while copied homework teaches you nothing. The most useful prompt is “here’s how I understood it: what am I getting wrong?”

At work, the most rewarding starting point is preparation: structuring a memo before writing it, anticipating objections before a meeting, rephrasing a delicate message, understanding a document you received the day before. The time saved is immediate, and the result stays under your control.

In everyday life, the most common needs are paperwork and writing: understanding an official letter, writing a complaint, translating a document, planning a trip. One-off situations where you don’t want to install an app or pay for a subscription just to ask one question.

• studies: get explanations and check your understanding, not the finished homework
• work: prepare, structure, anticipate objections
• personal: official letters, complaints, translations, trips
• in every case, the decision and the final version are yours

Check an answer before you rely on it

The main flaw of an AI isn’t that it makes mistakes; it’s that it makes them with the same confidence as when it’s right. Nothing sets a solid answer apart from an invented one: no hesitation in the tone, no cautious wording.

The risky cases are fairly predictable, though, so you can protect yourself without checking everything. Precise figures are one: an amount, a percentage, a date or a legal reference can be made up and still look credible. Sources are another: a cited book or article may not exist. And anything that changed recently escapes a model whose knowledge stops at a given date.

The practical rule fits in one sentence: the more precise and checkable a piece of information is, the more it deserves to be checked. An explanation is checked for consistency; a figure is checked at the source. You can also ask directly: “what are you sure about in this answer, and what are you less sure about?”

• a wrong answer sounds as confident as a right one
• always check figures, dates, amounts and cited references
• recent events escape the model: cross-check with an up-to-date source
• ask it to separate what it’s sure about from what it isn’t

What an AI can’t do

It doesn’t know what happened recently. Its knowledge stops at a certain date, and the free version doesn’t search the web. For news, current prices, regulations or opening hours, it isn’t the right source: it’s asked to say when something may have changed, but it can’t always tell what it doesn’t know.

It doesn’t know your particular situation. In law, taxes or medicine, it describes a general principle correctly, which helps you prepare an appointment or understand the vocabulary. But an answer that’s right in theory can be wrong for your specific case, and it has no way of knowing. It doesn’t replace a lawyer, a doctor or an accountant.

It doesn’t decide for you either, even when it sounds like it. It suggests options consistent with what you’ve told it, which isn’t the same as advice that accounts for what you haven’t: your history, your constraints, what you know about the other person. The final judgment is yours.

• news, prices, current rules: use an up-to-date source
• law, health, taxes: a general principle, never your personal case
• it doesn’t know your unspoken constraints: the decision stays yours
• it can’t sense the tone of a relationship or the human context of a situation

Using AI at work: a few precautions

The first question isn’t technical but contractual: does your employer or your client allow external tools on the documents involved? Many companies have a written policy, and a confidentiality agreement with a client also covers online services. Checking takes five minutes and avoids a real problem.

The second precaution is anonymization, and it solves most cases. When it’s the mechanism that matters (understanding a standard clause, structuring a memo, rephrasing a message), replace client names, amounts and references with neutral placeholders. The result is just as good, and you haven’t shared anything identifying.

The third concerns decisions about people. Screening job applications, evaluating colleagues or settling an internal dispute with an automated tool raises specific legal obligations and engages your responsibility. The tool can help you prepare an interview or build an assessment grid; it should never decide someone’s fate on its own.

• check your company’s policy and your confidentiality agreements
• anonymize names, amounts and references: the result is the same
• no automated decisions about people
• reread before sending: you remain the author of what you sign

Write a tricky message, and reread it before you send it

It’s one of the most useful everyday uses, and one of the least talked about. Replying to an unpleasant email, chasing a client who hasn’t paid, turning down a request without offending anyone, breaking bad news: these messages cost time and energy, less because they’re long than because you hesitate over the tone.

What works best is to write your own version first, even clumsy, even too blunt, then ask for a rewrite with a precise instruction about tone. “Rewrite this so it stays firm on substance but courteous in form, for a client I want to keep working with” gives a usable result, where “write me a message” produces a template unrelated to your situation.

The next step is the one everyone forgets: have the message checked before you send it. Ask “how could this be misread by someone who reads it quickly and in a bad mood?” That question brings out ambiguities you no longer see, because you know your intention, not your reader. And for a message written in anger: write it, have it rephrased, then wait an hour before sending.

• write your version first, then have it rephrased
• specify the tone and your relationship with the recipient
• ask how the message could be misread
• always reread: you remain the author of what you send

Create an image instead of text

Not every need is written. An illustration for an article, a mood image for a presentation, a thumbnail for social media or simply a visual idea to explore: describe it in the chat and the assistant generates the image. Image generation is mainly a Premium feature; the free version offers a small trial.

The quality of a generated image depends almost entirely on how precise the description is. Three things change everything: the subject, the visual style and the framing. “A cat” gives you an ordinary image; “a ginger cat asleep on a windowsill, late-afternoon light, photograph, close-up” gives you a usable one. The style you ask for (photo, illustration, watercolor, technical drawing) weighs as much as the subject.

Two limits are worth knowing. Text inside images can still come out approximate: don’t count on a poster with a perfectly spelled slogan. And don’t generate the faces of real people. Before publishing an image, especially commercially, check the usage rights: rules differ between tools and change regularly.

• describe the subject, the visual style and the framing
• the style weighs as much as the content
• text inside images can be approximate: avoid slogans and posters
• mainly a Premium feature; the free version offers a small trial

Get your own writing proofread and corrected

Having your own text corrected works better than having it written for you, for a simple reason: the text keeps your voice. A correction covers spelling, grammar and clarity; a full rewrite produces a smooth, often recognizable text that doesn’t quite say what you meant.

Say exactly what you want corrected, or everything will be reworked, including what was simply your style. “Fix only spelling and grammar, don’t touch the phrasing” is a very different instruction from “improve this text”: the second one hands you back someone else’s text.

To improve rather than just correct, ask for explanations: “correct it, then explain each correction in one sentence.” Your recurring mistakes (agreement, punctuation, repetition) show up quickly, and spotting them is how you stop making them. For a text others will read, one last question helps: “what isn’t clear to someone who doesn’t know the subject?”

• a correction keeps your voice, a rewrite replaces it
• define the scope: spelling only, or style too
• ask for an explanation of each correction to improve
• have it point out what’s unclear to a newcomer

An independent alternative to ChatGPT

ChatGPT is OpenAI’s product. Chat Français is a separate, independent project based in France, with no connection to OpenAI. Free conversations run on Qwen, a language model developed by Alibaba Cloud. Premium subscribers are answered by an OpenAI GPT model, accessed through OpenAI’s commercial API, and image generation relies on OpenAI too.

Because the service is run from France, it’s subject to the EU’s GDPR. The legal pages and the privacy policy are in French for now; the privacy policy lists every provider that processes your messages. As with any online chatbot, don’t paste passwords, bank details or sensitive medical information.

Without an account, there’s no conversation history: your chat is tied to an anonymous session in your browser, and we have neither your name nor your email address. With a free account, your conversations are kept and you can find them on all your devices.

• an independent project based in France, not affiliated with OpenAI
• free chats run on Qwen (Alibaba Cloud), Premium on an OpenAI GPT model
• subject to the EU’s GDPR; the privacy policy lists every provider
• never paste passwords, bank details or sensitive medical information