How ChatGPT tokenization works
ChatGPT runs on OpenAI GPT models, so its tokenization is identical to the OpenAI API: about 4 English characters or 0.75 words per token.
Long conversations matter: ChatGPT re-reads recent conversation history on every turn, so earlier messages consume context window alongside your new message.
If a pasted document exceeds the context window, ChatGPT silently loses the oldest content. Counting first tells you whether to split the document.
Everything you paste here stays in your browser, which makes this safer than testing sensitive text inside ChatGPT itself.