Conversational forms, and when they are worth it.

A conversational form is a form that asks its questions one at a time, as a chat, instead of showing every field on one page. The newer kind goes a step further: it reads each answer as it arrives, asks a follow-up when the answer is too thin to use, and answers the respondent’s own questions before carrying on. Both kinds are sold under the same name, and they are not the same product.

The two things “conversational form” means

Search for a conversational form builder and you will find two categories filed under one label.

The first is presentation. One question per screen, a progress bar, a friendly tone, Enter to continue. Typeform made this style famous, and WordPress plugins such as WPForms, Gravity Forms and Fluent Forms now offer it as an add-on. The form is exactly as fixed as before; it is just revealed a piece at a time. That alone helps, because a visitor judges a form by how long it looks, and one question looks short.

The second reads the answers. An AI sits between the question and the record. When somebody types “growth” into “what are you hoping to get out of this?”, it asks which kind of growth, and records the reply as the answer. When they ask “is parking included?” halfway through, it answers from what you told it and returns to the question it was on. This is what chatform is, and what Typeform sells separately as Formless and Jotform as AI Agents.

What changes for the person answering

Picture a new-client intake form: twenty fields on one page. The visitor scrolls to the bottom before typing anything, sees the budget question and the free-text box labelled “describe your project”, and leaves. Nothing about any single question put them off. The whole stack did.

Baymard Institute’s checkout research puts a number on that for shopping: 17% of US online shoppers have abandoned an order because the checkout was too long or complicated. A conversation spends that cost in small pieces, so the decision to leave gets made one question at a time instead of once, at the door.

The quality of the answers changes too, and that is the part people underestimate. A text box asks for effort and gives nothing back, so people write the least that will let them continue. Survey researchers call it satisficing. In a 2019 CHI study, Kim, Lee and Gweon found people answering through a chat interface satisficed less and gave more differentiated answers than people answering the same survey on the web. And in a 600-person experiment published in 2020, Xiao and colleagues found a chatbot that probed thin open-ended answers got responses that were significantly more informative, relevant, specific and clear than an ordinary Qualtrics survey.

A third effect is older than chatbots. Schober and Conrad showed in the 1990s that when an interviewer is allowed to explain what a question means, answers get much more accurate. A form cannot explain itself. A conversation can, which is why the ability to answer the respondent’s questions matters as much as asking better ones. We keep the full set of studies, with DOIs, on why conversation works.

When a normal form is the better choice

This part gets left out of most pages selling conversational forms, so here it is plainly. A newsletter signup, a login, a three-field contact form where every field is obvious: use a normal form. A conversation adds turns there and collects nothing extra.

Conversation is also slower. Conrad and Schober’s follow-up household telephone survey found clarifying questions improved accuracy at the cost of longer interviews. That is a good trade for an intake form, a job application or a feedback survey, where a vague answer costs you a follow-up email later. It is a bad trade anywhere speed is the whole point.

A rough rule we use: if any question on the form could reasonably be answered with “it depends”, or you have ever emailed someone back to ask what they meant, it is a conversation. If not, it is a form.

What a real form taught us

In September 2026 a university hackathon ran its team registration on chatform: nine questions covering the team, the problem statement it was taking on, the leader’s contact details, the other members, an eligibility check and a declaration, with sign-in required before the first one so each student could register only once. Over the week it was open, 129 people started a registration and 55 finished it, a 43% completion rate for a form that needed a whole team’s details from one person.

Two things in that data were more useful than the headline number. First, most of the people who never registered were lost before question one. Roughly three in four sessions ended without a single answer, on a form whose first step was a sign-in screen. (Sessions include reloads and repeat visits, so treat that as a direction rather than a precise rate.) No amount of good conversation fixes a door people will not walk through, and if you require sign-in, that is the first place to look.

Second, 98 people who left partway were sent at least one reminder, and 24 of them finished after it arrived. We do not call that “24 recovered”, because the form ran without a holdout group and some of those students were always coming back before the deadline. The research that does hold up is narrower: Sauermann and Roach found that reminders which change their wording raised the odds of a response by over 30% against reminders that repeat themselves. That is why chatform’s reminders (on the Pro plan and above) are worded differently each time, and why it can hold a share of people back so the number you read has a control behind it.

What to look for in a conversational form builder

Whichever product you pick, these are the questions worth asking before you commit:

• Does it follow up on thin answers, or only display them nicely? This is the line between the two categories above, and the one that changes what you collect.
• Can the AI change the order or skip a required question? It should not be able to. In chatform the flow is owned by a state machine and the model can only act inside it; we wrote up why the model never owns the conversation.
• Is the branching checked before you publish? Conditional logic breaks quietly. A good builder refuses to publish a flow with a dead end.
• Are half-finished answers kept, and can you go back for them? Most tools either drop partial answers or send one reminder at most.
• What does it cost per response? Typeform’s free plan stops at 10 responses a month and Jotform’s at 100; chatform, Tally and Youform do not cap responses on the free plan. Our side-by-side comparison has the full table, with the date each price was read.
• Can you embed it where your visitors already are? Inline, as a popup or a side tab, or as a link in a WhatsApp message or a QR code.

How to build a conversational form

1. Start from what you need to know, not from fields. Write down the decisions the answers have to support (“can we take this job?”, “which session should they join?”), then the questions that settle them.
2. Describe it, or pick a template. In chatform you can paste a description and get a draft from the AI form builder, or start from one of the free form templates, each with its questions and branching already written.
3. Ask the open questions openly. “What should we know before the first session?” is where a follow-up on a thin answer pays off, so do not turn it into a dropdown. Choice and consent answers are matched exactly and never go to the model.
4. Give it something to answer questions from. Prices, opening hours, what happens next. Without it, the respondent’s question is a dead end. (In chatform, the knowledge base starts on the Pro plan.)
5. Share it and read the drop-off. Per-question drop-off tells you which question is losing people; fix that one first.

If you want the whole job for one situation, the guides by use case walk through bookings, quotes, intake, feedback and hiring end to end.