Building forms
Follow-up questions Premium
Dig deeper into an answer with a follow-up question, asked when an answer is vague or when it matches a condition you set.
"Dig deeper" asks a respondent one or two short follow-up questions about the answer they just gave, before moving on. A follow-up is the question a good interviewer would ask next: "You mentioned checkout was confusing. Which step lost you?"
Two ways a follow-up is triggered
When the answer matches a condition. For example: when a rating is below 3, or an NPS score is below 7, ask what went wrong. You set the condition exactly like a logic condition, so deciding costs nothing and happens instantly. This works on any question that offers comparisons: text, numbers, ratings, scales, choices, yes/no, dates and times.
When the answer is too thin to be useful. For short and long text, an AI model reads the answer and asks a follow-up only when it's vague, such as "it was fine" or "pricing". A specific answer moves straight on.
On a text question you can choose either. Setting a condition replaces the judgement: every answer that matches it gets a follow-up, vague or not.
Setting it up
Select a question, open Dig deeper in its settings and choose Turn on.
| Setting | What it does |
|---|---|
| Dig deeper when | The condition that triggers a follow-up. Required for questions other than text. |
| First follow-up | Your own wording for the first follow-up, such as "What's the main thing we'd need to fix?". Optional; can quote answers with Insert answer. |
| What makes a good answer here? | For judged text answers: tells the model what a useful answer looks like, which steers both the judgement and the follow-up's wording. |
| How far to dig | One follow-up, or Up to 2 follow-ups. |
If you write a First follow-up, it appears instantly and needs no AI at all. If you leave it blank, the model writes one based on the respondent's answer. A second follow-up is always written by the model, and only if the first reply is itself thin.
A single response gets at most 4 follow-ups in total, however many questions have them turned on.
What respondents see
The follow-up replaces the question on screen, typed out as if someone were asking it, with a text box, OK and Skip. It keeps the question's place in the counter and progress bar, since it's part of the same question.
Follow-ups are never required. Skipping, or pressing OK with nothing typed, simply moves on, and a respondent who skips one isn't asked another about the same answer.
What follow-ups never do
- They never change where the respondent goes next. Logic only looks at answers to questions, never at follow-up replies.
- They never hold a respondent up. If the AI is slow, unavailable, or declines, the form moves on as if follow-ups were off.
- They never run inside a question group or on a form with All questions on one page, where questions are answered together.
- They never apply to questions a respondent didn't end up on. If someone goes back and changes an answer, its follow-ups are cleared, and they're dropped with answers from any path the respondent left.
Reading the replies
Follow-ups and their replies are stored with the response, under the answer they're about:
- In the response drawer, they're indented under the answer.
- In CSV exports, each follow-up gets two columns: the question asked and the reply.
- In webhooks, they're in the answer's
followUpslist. - AI insights and response summaries read them too.
Both the question and the reply are encrypted when the form is, since a good follow-up quotes the answer it came from.