How to prompt AI for better survey answers

AI prompts for surveys can help you generate insightful questions that deliver results.

However, the prompts you give to an AI platform like ChatGPT must be precise and carefully considered.

Otherwise the questions it generates may lack the depth and relevance you need.

We’ve spent months researching how to generate better AI prompts for the release of our AI Survey Creation feature.

We’ve concluded there are 5 main categories to consider when your task is more complex.

1. Role
2. Task
3. Context
4. Constraints
5. Output

Not every task will require this, but it’s good to bear in mind.

Let us explain.

1. Role Tell the AI which perspective they should speak from

One of the biggest mistakes people make when asking AI to create a survey is simply saying:

“Act as a survey expert and generate some questions.”

While the questions you get may be adequate, a more specific prompt can+ help you get questions that dig deeper and lead to greater insights.

The quality of the questions depends heavily on the perspective you give the AI.

For stronger survey questions, ask AI to combine several perspectives.

2. Task: Give precise instructions

Vague instructions leads to a vague response – so be specific.

In our research, we’ve found that using the following words in a prompt work well.

• Create
• Summarise
• Compare
• Explain
• Rewrite
• Analyse
• Categorise
• Suggest
• Critique
• Extract
• Transform

These are strong words that lead to action and better outcomes.

For example:

This is a much better prompt than simply asking AI what it thinks about two question options.

When making your prompt, consider if you would expect a colleague to know exactly what you were looking for with the same information.

3. Context: What does AI need to know?

AI can only work with the information it has been given or can access. If background information is missing, AI can only make assumptions.

For example: “Create a survey for training courses ” is quite limited.

It is better to include context, such as who the training is for, their prior knowledge, what topics should be included, and what the outcomes should be.

Here’s a better prompt with some added context:

Now the AI has something to work with.

General context

Some key things to consider for adding context to your prompts include:

• Who the audience is
• What they already know
• What the purpose is
• What has already happened
• Relevant background information
• Examples of existing work
• The problem you are trying to solve

If you’re unsure, always come back to the key things AI needs to know to better assist you.

Because that is the useful context you need to include.

4. Constraints: What rules should the AI follow?

Without constraints, AI tends to decide things for itself, so you should add constraints and parameters to lead AI to the quality answers you’re looking for.

A simple prompt about creating 5 customer satisfaction survey questions” will give limited question types.

A better approach would be adding constraints to your prompt.

Such as:

This works because you’ve reduced the number of decisions AI has to make by itself.

Some Constraints that we’ve found to be effective include:

• Length
• Number of items
• Tone
• Language
• Structure
• Required information
• Things to avoid
• Formatting

5. Give AI an example

One of the easiest ways to improve the output is to show AI exactly what you’re looking for

If you already have an example of a survey question, report, table or document with the style you want, include it in your prompt and say something like:

This is often much more effective than trying to describe your preferred format in several paragraphs.

Don’t just ask for questions – be specific.

For example, instead of:

Try:

This ensures AI knows what to create, how to structure it and how it should sound.

By specifying the output, you can save time and get a much better result.