New

Now in Claude, ChatGPT, Cursor & more with our MCP server

Back to docs
Comparisons

How to Convert a Typeform Survey Into an AI Interview

A step-by-step migration guide. Map every Typeform question type to Koji structured questions, get AI-moderated probing, voice or text mode, and a real-time aggregated report.

The Bottom Line

If your Typeform responses feel shallow, the form itself is not the problem — the medium is. A survey can ask "What did you think?" but cannot ask "Can you tell me more about that?" The fastest fix is to convert the survey into an AI-moderated interview, where the same structured questions still chart correctly and every answer can be probed for depth. With a tool like Koji, this migration takes about 20 minutes per survey: map each Typeform field to one of Koji’s 6 structured question types, set probing depth, and publish the interview link.

This article is the field-by-field migration guide.

Why Migrate at All?

Typeform is great at the things surveys are great at: capture, branching, and aesthetic forms. The structural ceiling is that a survey response is a single answer per question. There is no follow-up. There is no "wait, what did you mean by that?" There is no themed analysis across responses.

Migration unlocks four things the form cannot:

  1. AI follow-up probing — every question can be configured to probe up to three follow-ups
  2. Voice OR text mode — participants choose; voice answers feel like a real conversation
  3. Real-time themed analysis — open-ended responses are auto-themed across participants
  4. Per-question structured aggregation — scale, choice, ranking, and yes/no questions auto-chart

The quantitative part of Typeform stays — the open-ended part gets a 10× upgrade.

The Migration Map: Typeform Field → Koji Question Type

Koji has six structured question types: open_ended, scale, single_choice, multiple_choice, ranking, and yes_no. Every Typeform field maps cleanly to one of them.

Typeform fieldKoji typeNotes
Long textopen_endedSet maxFollowUps to 1 or 2 for depth
Short textopen_endedSet maxFollowUps to 0 if you want it tight
Multiple choice (one answer)single_choiceUse config.options for the choices
Multiple choice (many answers)multiple_choiceUse config.options, optionally allowOther
Yes/Noyes_noRenders as buttons in text mode
Opinion scale (1–5, 1–10)scaleUse scaleMin, scaleMax, optional scaleLabels
Net Promoter Score (NPS)scalescaleMin: 0, scaleMax: 10, anchored labels
Rating (stars)scaleSame as opinion scale
RankingrankingDrag-to-reorder widget in text mode
Picture choicesingle_choice or multiple_choiceImage rendering depends on Koji widget version
Dropdownsingle_choiceTreated as a long option list
Email / Phone / NumberIntake form fieldMove to the optional lead form before the interview
File uploadNot supported in interview flowUse a follow-up email
Statement (no input)Skip in question listUse the intake screen description field instead
Question groupUse section fieldGroup questions logically in the report

Step-by-Step: Convert in 6 Steps

Step 1 — Export your Typeform questions

In Typeform, go to Create → Logic / Content and copy your question list. Note the type, the answer options, and any logic rules. Save this somewhere you can paste from.

Step 2 — Move identifying fields to the intake form

Email, name, phone, and demographic short-text fields belong on Koji’s intake screen, not in the interview itself. The intake form supports text, email, phone, select, textarea, and checkbox field types — same coverage as a Typeform identifier block. This keeps the interview itself focused on insight questions and prevents identifying fields from consuming probing follow-ups.

Step 3 — Create the Koji study

In Koji, create a new study. Set:

  • Headline + description on the intake screen (use Typeform’s welcome screen text here)
  • interactionMode to allow both voice and text, with voice as default
  • defaultLanguage if your Typeform was in a non-English locale
  • Brand colors and orb color to roughly match your Typeform aesthetic

Step 4 — Map each question, one by one

For each Typeform question, create a Koji structured question with the matching type. Three details matter:

  • required: true for any Typeform field that was required
  • config.options must match Typeform’s option list exactly if you want before/after data comparability
  • probing.maxFollowUps is the magic ingredient — set it to 0 for screener-like questions, 1 for standard depth, 2–3 for the questions you most want a real conversation on

For open_ended questions specifically, write probing.instructions to guide what the AI should probe on. Example: "If they mention a specific competitor, ask which features they used. If they mention pricing, ask what they compared it to." This replaces the manual moderator instinct.

Step 5 — Replicate Typeform branching with sections and required flags

Typeform branching maps loosely to Koji’s section field plus required: false flags. Koji’s AI moderator handles natural conversational branching automatically — if a participant indicates they have not used the product, the moderator will skip product-specific follow-ups. For complex hard-branching logic (e.g., pricing-only questions for finance leaders), gate them via the screener instead.

Step 6 — Publish and migrate the link

Replace your Typeform URL with the new Koji interview link wherever it lives — email signature, in-product, post-purchase, NPS routing. Watch the report populate in real time.

What You Gain

A before-and-after comparison for a typical 8-question Typeform:

CapabilityTypeformKoji AI interview
Long-text follow-up probingNoneUp to 3 follow-ups per question
Voice modeNoYes (voice + text)
Multilingual (single link)Limited30+ languages, auto-detected
Auto-themed open-ended analysisNoYes
Per-question quantitative aggregationBasicYes, auto-charted
Quality gate on responsesNoYes (only score 3+ counts)
Real-time reportPartialYes
Cost per response€0 (Typeform) + analyst hours€1 text / €3 voice, analysis included

The meaningful unlock is the probing depth. On a typical churn or NPS-detractor question, the qualitative follow-ups produce 3–5× more usable insight per participant than the equivalent Typeform long-text field — with no extra researcher time.

What Stays the Same

Four things should not change in the migration:

  1. Question wording — keep the wording identical so you can compare data before and after.
  2. Required flags — preserve which questions are mandatory.
  3. Option labels — for single_choice and multiple_choice, match Typeform exactly.
  4. Scale ranges — if Typeform used 1–10 NPS, Koji should use scaleMin: 0, scaleMax: 10 (not 1–10 — NPS starts at 0).

This preserves data continuity and lets you do a clean A/B between the survey-era and interview-era results.

Common Questions to Plan For

"Will participants drop off because interviews feel longer?"

Async AI interviews tend to increase completion rates compared to long-form surveys, because the conversational pacing feels less like work. The biggest dropoff risk is at the intake screen — keep it short.

"What about anonymous responses?"

Koji supports fully anonymous interviews (no intake fields required, optional anonymity badge on the intake screen). The voice mode does record audio for transcription; document this in your consent footer.

"How does this affect my GDPR posture?"

Both voice and text interviews generate transcripts that you control. Configure the consent footer with your DPA/privacy link, set retention policy on the workspace, and you are operating on the same legal basis as a recorded user interview.

"Will I lose Typeform’s logic / branching?"

Most Typeform branching collapses into a) the AI moderator’s natural conversational branching and b) Koji’s screener flow. For very complex branching, you can run two parallel Koji studies linked from a single intake.

When To Skip the Migration

Keep Typeform for:

  • Short signup or contact forms with no insight questions
  • Static lead-capture flows that feed a CRM and never need follow-up
  • High-volume public polls where depth is not the goal

For anything where you actually want to learn — feedback, NPS, churn, discovery, feature validation, willingness-to-pay — the interview format wins.

Related Resources