When it comes to prompting, there are two schools of thought. Some experts favour long, highly detailed prompts, which can be very effective. The problem is that they do not sit easily with the day-to-day reality of a journalist’s work: fast-moving, time-poor, and often shaped by uneven levels of AI know-how. A second approach backs very simple prompts. They may be less powerful, but they are also more practical — and easier to remember.

Here are a few examples.

/realtalk. Start your request with the command /realtalk. ChatGPT’s reply will sound much closer to natural speech.

/simpleten. Start your request with the command /simpleten. ChatGPT’s reply will be genuinely clear and basic.

/nofluff. Start your request with the command /nofluff. ChatGPT’s reply will be lean, practical and free of filler.

/prolevel. Start your request with the command /prolevel. ChatGPT’s reply will read like it comes from a true subjectmatter expert — in tone, terminology and the robustness of its content.