Separate your instruction from the text it acts on

prompting basics

// what it does

Put the instruction at the top and wrap the pasted material in delimiters like triple quotes or an XML-style tag. This tells ChatGPT exactly where your command ends and the raw content begins, so it does not mistake a sentence inside your document for a new instruction. OpenAI's own best-practices guide leads with this tactic.

// prompt

> Summarize the text below as a bullet-point list of the most important points. Text: """[paste text]"""
> Translate the text between the tags into French; keep names unchanged. <text>[paste text]</text>

// gotcha

Without delimiters, instructions buried under a wall of pasted text are easy for the model to overlook or misread — especially if the text itself contains something that looks like a command. Lead with the instruction, then delimit the content.

// resources