A language model answers the question it was actually asked, not the one you had in mind. Prompting well is mostly about closing that gap.
Seven techniques that work
State the task and the goal
"Summarise this report for a manager who has two minutes" works better than "summarise this".
Give context
Who the answer is for, what you already know, and what you'll use it for.
Show examples
One or two examples of the output you want ("few-shot prompting") are often clearer than a long description.
Specify the format
A table, a list of five points, a 100-word paragraph, valid JSON.
Ask for steps on hard problems
For calculations and reasoning, asking the model to work step by step, and to show the steps, makes errors easier to spot.
Set constraints
Length, tone, what to avoid, and what to do if information is missing ("say you don't know").
Iterate
Treat the first answer as a draft. Point out exactly what to change.
Always check
Models can state wrong facts confidently. Check numbers, names, dates and quotations against reliable sources, and never paste passwords or confidential data into a service you don't control.
Tools for this
Asaiejadoo — everyday calculation and AI guides, part of the ZIBADIS network founded by Masoud Moghaddam in Tehran.
