Why use synthetic interviews?

Let’s say your team hasn’t adopted AI for analysis yet, and you think they’re falling behind.

You really want to evaluate whether AI can actually handle analysis of the kind of complex conversation material you will have from an upcoming project.

The challenge is: you won’t be allowed to use AI for that project until you can prove to your manager that it’s valuable and worth the risks.

Synthetic data is a solution here.

You can run your own tests by generating synthetic interview scripts based on the kind of interview protocol and questions you’ll be using in that study. That way, you’ll have a look-alike analysis project you can use to show the team how AI worked.

</aside>

Generating synthetic interview transcripts with AI can help you:

✔️ Create realistic, role-specific interviews without recruiting participants

✔️ Explore different user personas and pain points quickly

✔️ Generate training data for AI-based insights or user research testing

My best tip to start off: 👉 Use Claude for this—it generates more natural and uninterrupted transcripts than ChatGPT.

Step 1: Define your scenario

Step 2: Find Good Interview Examples

Step 3: Use a Real Interview Script

Step 4: Generate synthetic interview data

Example output

Bonus: Prevent AI from stopping mid-generation

A few final thoughts: generating the best synthetic interviews

✅ Use Claude for more natural, varied responses

✅ Structure interviews with open-ended questions & realistic tangents

✅ Test few-shot prompting if AI responses feel repetitive or generic