AI Disclosure

Every reply on OpenChatGPT is generated by artificial intelligence. This page explains how, and what that means for you.

What generates the replies

When you send a message, it is forwarded by our server to a third-party AI model provider through an API. A large language model generates the reply. The site's gateway automatically picks the best available model for each question, so you always get a capable model without choosing anything. There is no model picker and nothing to configure.

No human writes the replies, and no human reviews them before you see them.

What OpenChatGPT is not

OpenChatGPT is an independent website. It is not ChatGPT, Claude, OpenAI, Anthropic, or Google, and it is not affiliated with, endorsed by, or sponsored by any of them. The models it uses are third-party models accessed through an API; using a model does not make this site that model's official product.

Replies can be wrong

AI models predict plausible text. They can:

• Invent facts, dates, quotations, and sources that do not exist;
• Give outdated information, because their training data has a cut-off date;
• Be confident while wrong, especially on specialist topics;
• Reflect biases present in their training data;
• Produce code that looks right but has bugs or security problems.

Verify important facts from a reliable source before acting on them — especially for health, legal, financial, and safety decisions. Do not use AI replies as a substitute for professional advice.

How your messages are handled

Your message is sent to our server, which forwards it to the AI provider to generate a reply. Your chat history is stored only in your own browser (localStorage). We do not keep an account database of your conversations. Do not type passwords, ID numbers, bank details, or other sensitive information into the chat. See the privacy policy for details.

Using replies responsibly

You may use replies for learning, drafting, and brainstorming. Check them before publishing or submitting them anywhere, and do not present AI-generated text as your own original work where that would be dishonest — for example, in exams or assessed work. You are responsible for how you use the replies.

How replies are generated, step by step

When you press Send, your message travels in three steps. First, your browser sends it to this site's server. Second, the server forwards it to a third-party AI provider through an API — the same kind of connection apps use to talk to AI models. Third, a large language model at that provider generates a reply, word by word, and the server streams it back to your browser, so you see the answer appear as it is written. There is no model picker and nothing to configure: the site's gateway automatically picks the best available model for each question. If the main provider is temporarily unavailable, the server falls back to a second provider, so the chat keeps working. Your conversation history for the current chat is sent along with each new message, so the model can answer follow-up questions in context — but that history lives only in your own browser, not in an account database on this site.

What the AI cannot do

It helps to know the edges. The AI can only see what you type or attach in the chat — it cannot see your files, your screen, your emails, or anything else on your device. It cannot run code, open a program, or check a website for you; it can only write text, including code for you to run yourself. It cannot remember you between visits: when you close the browser or clear your history, the conversation is gone, and a new visit starts fresh. It does not know today's news, the current price of anything, or events after its training data was collected, so it can give outdated answers about anything time-sensitive — always check live facts (prices, dates, opening hours) from a current source. And it has no real understanding: it predicts plausible text, which is why it can sound confident about something that is simply wrong.

Checking facts and using replies

Treat every reply as a first draft, not a final answer. For everyday questions — an idea, a summary, a first version of a text — that is usually enough. For anything important, verify before you act: check dates, names, numbers, and quotations against a reliable source, and never rely on an AI reply alone for health, legal, financial, or safety decisions. If you use a reply for work or study, read it first, fix anything wrong, and make it your own — do not submit AI-generated text as your own original work where that would be dishonest, for example in exams or assessed assignments. Code needs the same treatment: read it, run it, and test it, especially the edge cases, before you ship it or hand it in. Used this way — as a fast, tireless drafting and explaining helper, with you doing the checking — AI chat is genuinely useful. Used blindly, it is a fast way to be confidently wrong.

Questions

Anything unclear? Use the contact page.