Previewing Your Interview Before You Share It

To preview a Koji interview before you share it, publish your study, open its Interviews page and go to the Preview tab. Choose Text or Voice, pick a language, and click Start text preview or Start voice preview. You then take the interview yourself, exactly as the AI interviewer will run it with participants, including its follow-up questions and your structured questions. Previews are free: they never use credits, never count as responses and never appear in your reports. You can run up to 20 text and 3 voice previews per study each day, and read back every preview afterwards.

This guide explains how previews work, what they include and leave out, and how to use them to pressure-test a study before real participants see it.

Why preview an interview

Every research method has a pilot step for a reason. Questions that look clear on the page can confuse in conversation, a follow-up can go somewhere you didn't expect, and a study can run longer than you planned.

Pilot testing also catches less than people assume when it's done lightly. In "Sample size for pre-tests of questionnaires" (Quality of Life Research, 2015), Thomas Perneger and colleagues showed that with 10 pre-test participants, a problem affecting 5% of respondents is detected only 40% of the time; with 20 participants, 64%. Their conclusion: the 5 to 15 participants common in pre-tests may miss even common problems.

With traditional surveys, you preview a static form. With an AI-moderated interview, the conversation adapts to each answer, so the only real test is to have the conversation. Koji's Preview tab makes that free and repeatable: run it several times, answer differently each time, and see how the interviewer responds before a single credit is spent.

Step by step: running a preview

Step 1: Publish your study

Previews run the real interviewer, which is created when you publish. On a draft study the start button is disabled with a reminder to publish first. If you've just added voice interviews, republish before previewing by voice.

Previews still work after a study is paused or closed, so you can keep testing changes between rounds.

Step 2: Open the Preview tab

Open the study, go to Interviews, and choose the Preview tab. A badge on the tab shows how many previews you've saved.

Step 3: Choose text or voice

Under Run the interview yourself, pick a mode:

• Text: type your answers in a chat
• Voice: talk with the interviewer out loud, using your microphone

Each card shows how many previews you have left today.

Step 4: Choose a language and add your name

Pick one of the languages enabled for your study. This is how you check a multilingual study in each language before launch. Optionally add Your name, which appears in the transcript so teammates can tell whose preview is whose.

Step 5: Start and finish the interview

Click Start text preview or Start voice preview. The interview runs right there on the page. A banner reminds you it's a preview: "Not counted, not charged." Click End to stop early, or finish naturally and you'll see "Preview complete."

What a preview includes and what it skips

Previews go straight into the interview, so they're focused on the conversation itself.

Included

• The full AI-moderated interview, with follow-up questions
• Your structured questions, shown the way participants see them
• The language you choose
• Your study's look, including its brand theme if you use one

Skipped

• Screening questions
• The welcome screen, pre-interview form and consent step
• The participant feedback prompt and thank-you screen

To check the welcome screen and screener, open your interview link yourself in a private window. Keep in mind that a real run through the link is a real response.

Previews are free and stay out of your results

Previews:

• Never use credits. They're marked as not billed from the start.
• Never count as responses. They don't appear in your response list or count toward response limits.
• Never appear in reports or insights. Report generation, insights and exports ignore them.
• Are never analysed. There's no quality score or summary for a preview.

That means you can test as much as your daily limit allows without worrying about polluting your data.

Daily limits

Each study allows 20 text previews and 3 voice previews per day. Limits reset at midnight UTC. Deleting a preview doesn't give it back, so the count reflects how many you started that day.

Reading back past previews

Under Past previews, each preview shows who ran it, its status (Completed, In progress or Ended early), text or voice, the language, the date and the duration. Click Read transcript to see the full conversation with timestamps, or Resume to return to one that's still running.

Previews made before your latest publish are tagged before republish, so you can compare the interview before and after a change.

From the row menu you can open a preview in a new tab, Copy link to share it with a teammate, or Delete it. Clear all removes every preview on the study; real responses aren't affected. Otherwise previews are kept until you clear them.

Who can run previews

Anyone with edit access to the study can run, end and delete previews. People with view access can see the Preview tab and read transcripts but can't start one.

Voice previews need a plan that includes voice interviews, which means Interviews or Enterprise. Text previews work on every plan, Free included.

How to get the most from a preview

• Play three participants. Run one preview as an enthusiastic customer, one as a sceptic and one who gives short answers. Watch how the interviewer probes each.
• Check the timing. Compare the preview's duration with the length you promised on your welcome screen.
• Test every language. Run at least one preview in each language you've enabled.
• Test voice on a real device. Use the laptop or phone your participants are likely to use.
• Republish, then preview again. The before republish tag makes it easy to see whether your edit fixed the problem.
• Share transcripts for sign-off. Copy a preview link to a stakeholder instead of describing the interview in a meeting.

For deeper pilot methodology, see Pilot Study in User Research and Cognitive Interviews.

Preview first, then pilot with real people

A preview tells you how the interview behaves. It can't tell you how your audience will react, because you already know what you meant by every question. Use both steps.

1. Preview until the interview runs cleanly. Fix confusing wording, trim questions that run long and check that each structured question appears where you expect.
2. Pilot with a small batch of real participants. Share the link with a handful of people from your target audience. Their responses are real, so they use credits and appear in your results, which is the point: you see the analysis and report on genuine answers.
3. Read the pilot transcripts before scaling up. If participants misread a question, edit it, republish and preview again.

Because Koji's AI interviewer moderates every conversation and analyses responses automatically, the jump from pilot to full launch is just sharing the link more widely. There's no moderator schedule to rebook.

Related Resources

• Structured Questions in AI Interviews
• Publishing Your Study
• Pilot Study in User Research
• Voice vs Text Interviews
• Text Interview Experience
• Voice Interview Experience