Build a form with AI
Just describe what you want — your AI asks a couple of simple questions and builds it.
You don't design a form field by field. You say what you want, in plain language, and your
AI client — talking to FormMaker's create_hosted_form tool — asks a couple of simple
questions back before it builds anything.
How it goes
- You describe the goal. Not fields, not field types — what the form is for.
- The AI asks one simple question at a time. Plain language, no jargon, no dumped checklist of technical options.
- It confirms in a sentence. So you can correct it before anything's built, not after.
- It builds. You get a hosted form, ready to share or embed.
This works for a trivial three-field contact form — the AI usually just builds it — and for something with real structure behind it, like a scored assessment.
Worked example: a scored assessment
Build me a form that scores how conversion-ready someone's forms are.
That's a goal, not a spec. A good AI client won't guess at the scoring — it'll ask, one question at a time:
- "What are you assessing?" — you answer, e.g. field count, question order, CTA copy.
- "What makes an answer strong or weak?" — so each option can carry the right points.
- "What should the result say for a high score vs. a low one?" — the copy for each outcome bucket.
Each answer gets a one-sentence confirmation before the next question. Once it has enough,
it builds the form: fields with per-option scores, and result buckets ("just getting
started" vs. "well on your way") that show after submit — the same optionScores and
results shape behind create_hosted_form.
You never see that shape. You answered three plain questions.
It's the same flow either way
A one-field newsletter signup and a ten-question scored assessment go through the same loop — describe, answer a question or two, confirm, build. The difference is just how many questions it takes to have enough to build correctly.
See also
- Create a form — starting from a template or from scratch.
create_hosted_form— the full tool reference, includingoptionScoresandresults.- Templates — ready-made starters, if you'd rather point at a field set than describe one from scratch.