Prompting basics: how to ask AI for what you actually want

Most disappointing AI results start with a disappointing request. You type “write a welcome email for my membership,” get back something bland and over-excited, and conclude that AI writing is not for you. But the tool did exactly what you asked. It just had to guess at everything you did not say.

A prompt is simply the instruction you give an AI assistant. Learning to write a good one is the single most useful AI skill a membership site owner can build, and it takes minutes, not months. This guide gives you a simple five-part structure and shows it working on real membership tasks.

Why vague prompts get generic answers

An AI assistant fills every gap in your request with the most typical answer it can produce. Ask for a welcome email with no other detail and you get the average welcome email: cheerful, generic, and written for nobody in particular. The more of the gaps you fill, the less it has to guess, and the more the result sounds like it belongs to your membership.

You do not need special phrases or tricks. You need to say what a thoughtful human assistant would ask you before starting the job.

The five ingredients of a useful prompt

1. Role: who the AI should act as. “You are an experienced editor of beginner cooking lessons” sets a level of expertise and a point of view.
2. Task: exactly what you want done, starting with a clear verb. Write, summarize, rewrite, list, compare.
3. Context: the background it cannot know. Who the members are, what they already know, why this piece exists.
4. Format: the shape of the answer. Length, headings or bullets, reading level, what to leave out.
5. Examples: a sample of what good looks like, such as an email you wrote that members loved.

Not every prompt needs all five. A quick question may need only a task. But when a result disappoints you, the missing ingredient is usually one of these.

Before and after: a welcome email

Lena runs a sourdough baking membership. Her first attempt looked like this:

Write a welcome email for my baking membership.

The result opened with “Welcome to the family!”, used three exclamation marks, and promised “exclusive content” without saying what it was. Here is her improved version:

You are a friendly, practical baking teacher. Write a welcome email for new members of [membership name], a membership that teaches home bakers to make sourdough bread. Most new members have tried sourdough once or twice and had flat, dense loaves. The email should reassure them this is normal, tell them to start with [name of first lesson], and invite them to post a photo of their first loaf in [community space]. Keep it under 180 words, warm but calm, with no exclamation marks and no marketing phrases. Here is an email I wrote that sounds like me: [paste a past email]

The second prompt produced a short email that named the real frustration, pointed to one lesson and asked for one action. It still needed a light edit, but it was recognizably hers.

Before and after: discussion questions

The same approach works for smaller jobs. Compare these two requests for questions to post under a lesson.

Before: “Give me some discussion questions about starters.”

After:

Here is a lesson from my membership for home bakers: [paste lesson text]. Write five discussion questions I can post underneath it. Each question should invite members to share their own experience, not test their knowledge. Avoid yes or no questions. Make at least one question easy enough for someone who has not baked yet. Return a numbered list with no introduction.

The first version produces quiz questions. The second produces prompts members can actually answer from their own kitchens, because it says what the questions are for.

Your second message matters as much as the first

Treat a prompt as the start of a conversation, not a single order. If the first answer is close but not right, tell the assistant what to change instead of starting again:

• “Shorter, and cut the first paragraph.”
• “Less formal. Write it the way you would say it to a friend.”
• “Number three is the one I like. Give me four more in that style.”
• “You have assumed members are beginners. Most have been baking for years.”

Specific feedback works better than “make it better,” which only invites a different guess. When a conversation produces something good, copy the final version of your prompt somewhere safe so you can reuse it.

Read everything before a member does

Even a great prompt does not guarantee a correct answer. AI tools can state things confidently that are simply wrong, from a baking temperature to the name of your own lesson. Before anything reaches members, read it as they would: check the facts, the links and lesson names, the promises it makes, and whether it still sounds like you. For more on setting realistic expectations, see what AI can and cannot do for a membership site owner.

Keep members' personal details out of your prompts too. You can describe your members in general terms without pasting anyone's name, email address or messages.

Try it now

• Take a prompt you have already used and check it against the five ingredients. Add the missing ones.
• Run the old and new versions side by side and compare the results.
• Write down one piece of follow-up feedback that improved an answer, so you remember to use it again.
• Keep a short example of your own writing handy to paste into prompts.
• Review every output yourself before it goes anywhere near a member.

If writing a prompt feels familiar, that is because it is the same skill as briefing freelancers so you get what you need: say who it is for, what done looks like and what to avoid. And if you have not settled on a tool yet, choosing your first AI assistant walks through how to test a few on your own tasks.