Guides

AI form-builder prompts: specify questions, options, and limits

Write a Thunder Forms prompt with the questions, answer types, optional fields, and page structure you need. Review the generated draft before sharing it.

Portrait of Aditya Singh KhichiAditya Singh KhichiMaker of ThunderFormsPublished Updated 5 min read

"A volunteer signup form" leaves several decisions open. You might need a phone number, weekday availability, or a question about a particular activity. Name those requirements in the prompt.

Start with questions you already know you need. Thunder Forms applies field defaults, but the model still has to choose wording and structure from your instructions.

A prompt you can check

Create a volunteer signup form with four questions on one page. Require full name and an email address. Add an optional multi-select for availability with Monday, Wednesday, and Saturday as the choices. Finish with an optional paragraph for access requirements. Use "Send signup" for the submit button.

Count the four questions first. Check the email field type, the three availability options, and both optional toggles before applying the draft. This prompt gives you decisions to verify in the builder instead of a general request to interpret.

Use your own wording. If the form must ask exactly "Which session can you attend?", include that sentence and the available sessions rather than asking the model to invent them.

Give counts and page boundaries when they matter

The generator is instructed to keep basic requests small and respect explicit question counts. Check the count yourself. Explicit instructions can still produce a draft with missing or extra questions, which is why a specific prompt helps you review the result as well as generate it.

For a multi-step application, name the pages and place questions within them, such as contact details on the first page followed by project details on the second. Sections group related questions within a page, so specify whether you want a new page or only a heading within the current page.

Describe the answers you need

For a quantity, specify whole numbers and a minimum. For a choice question, give the options and say whether someone can select several. A rating request should name its range, such as 1 to 5, because the range affects both the displayed control and accepted answers.

Need context? Ask for helper text. A dietary question might need an allergies paragraph alongside fixed food preferences. Review that wording for your audience before treating the generated form as ready.

Use a template when its questions already fit

An unchanged template request uses the stored configuration directly, preserving its questions and submit-button text so you can start from the form you inspected rather than a model's reconstruction of it. You can inspect those forms in the template gallery first.

A template request with additions preserves the original pages and appends generated pages for the new questions. Additions appear on new pages. Move them in the editor if you want each new question beside a related existing one, checking the page order again before sharing the form.

Turn a vague request into a buildable specification

A useful prompt answers five questions: who will fill out the form, what decision it supports, which answers are required, what answer types fit, and what should happen after submission. For example, replace "make a feedback form" with "Create a post-delivery survey for customers who have received an order. Ask one required five-choice satisfaction question, then an optional paragraph about the delivery. Do not ask for a phone number. Keep everything on one page."

That prompt names the respondent, the moment, the required scale, an optional comment, a privacy constraint, and a page limit. It still leaves the exact wording of the scale open, so inspect the draft and make the choices consistent before using it.

Review generated fields against the request

Match every requested question to a field. Confirm the answer type, options, required status, and page placement. Check that the form did not add sensitive details or collect information you did not ask for. Then read the questions as a respondent who has never seen your prompt.

If one field is wrong, decide whether the error came from an unclear instruction or an unsupported request. Make the prompt more specific for the next draft, or correct the field in the regular editor when only a small change is needed. The person publishing the form remains responsible for its questions and collection settings.

Frequently asked questions

Short answers to the questions every reader asks next.

Check that the prompt names the question and its expected answer. Add the wording or options you need and try again. An explicit request improves the instructions, but every generated draft still needs review.

Generation works at the form level. Template requests with additions preserve the template and append new pages. For an existing custom form, review how you apply a generated draft before replacing content, and use the regular editor for a small change.

Yes. Structural checks reject malformed generated forms, but they cannot determine whether the wording, question count, or answer choices match your intent. Preview every page and submit a test response.

In the builder, preview the generated form, check its constraints, and submit a test response before copying the public link.

AIForm builderPrompting

Keep reading

Guides6 min read
'Upload 1 supported file' blocks the attachment; the missing responses folder stops the form. What each error means, the fastest fix, and the route that avoids them.
Read article
Guides5 min read
Google Forms stays free and simple until a specific wall shows up. The four upgrade signals worth acting on, and the reasons to stay that outweigh them.
Read article
Guides7 min read
Collect resumes, contracts, receipts, and ID photos. See where files land, what limits apply in Google Forms and Microsoft Forms, and one route that needs no sign-in.
Read article