AI tools do what you ask them to, and if the request is generic, so is the result. A prompt such as “write text for my website” gives a bland text that would fit any company. If you describe who the site is for and how you are different, the result will be about your company.
A good prompt resembles a good brief. It contains the goal, the audience, the tone, the length and, if you have one, an example of something you like.
It helps a lot to give the AI a role, for example to answer as an experienced copywriter or UX designer, and to give it the rules you would tell a colleague. The more background material it gets, the less it has to guess.
What to put into a prompt
- Who the thing you are making is for, and what you want to achieve with it
- What tone it should take, and what it should avoid
- In what form you want the result, for example how many versions or how long
- Examples that show clearly what you have in mind
It is worth saving prompts that work well and reusing them as templates. In a team everyone then starts from the same place, the results become more consistent, and the template can be refined over time as it turns out what regularly needs fixing afterwards.
Prompting rarely works the first time. It is worth refining the result step by step and giving feedback on what is not right and what should be different. In a longer conversation the AI understands better and better what you want, so there is no point starting over every time. If a conversation has become too long and muddled, it is worth opening a new one with a summary.
The quality of the prompt is only part of the result. AI tends to give answers that seem convincing at first, yet look wrong or outdated to an expert. On a website this can be a button in the wrong place, in a text a claim that is not true of your company.
That is why it is worth having AI-made work reviewed before it goes public. In AI review we also show you how to rewrite your prompts so you get a better result in the next round.
