When you're writing a prompt for an AI tool, the model isn't actually counting your words — it's breaking your text into tokens, which are chunks that can be shorter or longer than a full word depending on the language model and the text itself.

A rough rule of thumb

For English text, a common approximation is that 100 tokens equal roughly 75 words, though this varies with punctuation, rare words, and formatting like code or JSON, which often tokenize less efficiently than plain prose.

Why this matters practically

• Context windows are measured in tokens, not words, so a "word count" estimate can be misleading
• Dense technical text or code can use up more tokens per word than casual prose
• Trimming a prompt's word count is still a reasonable proxy for reducing token usage, even without exact numbers

Start with word and character count

Without a live tokenizer handy, checking your prompt's word and character count with a simple counter still gives you a useful, fast estimate of roughly how much space your text will occupy before you send it to a model.