Design a study just by chatting

The clearer you are with the agent, the better the study it hands back.

The AI Study Designer turns a plain description of what you want to learn into a ready to review study draft. You answer a few questions in chat, it builds the screener and the activities, and you finish and launch in the Study Builder.

It works best when you meet it halfway. Here is how to get a strong study on the first try, followed by a full example from the first message to the finished draft.

Eight ways to get a great study

None of this is required. The agent will fill gaps with sensible defaults. But the more of this you bring, the less back and forth you need.

1. Say what you are testing and what you want to learn. “Find out why people abandon our checkout” beats “test my website.” A clear goal shapes every activity.
2. Be specific about the task. Name the exact action participants should attempt, and where it should stop.
3. Describe your ideal participant. The agent builds your screener from this. “US adults who shop online often” is enough to start.
4. Share your link when you have it. A website, prototype, or app store link lets the agent set up the task properly.
5. Use plain words. Write your study in sentences. Emojis and random symbols are not understood and slow things down.
6. One question at a time. Answer what it asks. You do not have to plan the whole study up front, though you can say more if you want.
7. Let it suggest, then adjust. Values marked “Suggested” are the agent’s best guess. Change any of them with a word.
8. It drafts, you launch. Nothing goes live from chat. You review, edit, and launch in the Study Builder.

A checkout study, start to finish

Meet a researcher at Trailhead, an online outdoor gear store. People fill their carts but drop off at checkout, and the team wants to know why. Here is the whole conversation.

Agent: Hi, I’m Userlytics’ AI chat. I’ll help you create your study. What do you want to do?

You: We run an online outdoor gear store, Trailhead. Lots of people fill their cart but never finish checkout. I want to find out where they get stuck.

Agent: Got it: an unmoderated test of your live checkout flow. What is the single most important thing you want to learn?

You: Whether the checkout steps are confusing, or ask for too much before people can pay.

Agent: Is there a specific task you would like participants to attempt?

You: Yes. Find a rain jacket, add it to the cart, and go through checkout up to the payment step.

Agent: What is the primary goal of this study?

You: Reduce drop-off at checkout.

Agent: Share the page you would like reviewed.

You: https://www.userlytics.com

Agent: How would you like to name this project?

You: Checkout drop-off, September.

Agent: Would you like to recruit from the Userlytics Panel, or bring your own participants?

You: Use your panel.

Agent: Describe your ideal participants and I will build a short screener from that.

You: US adults who shop online regularly and have bought outdoor or sporting goods in the last few months.

Agent: Perfect. I will target the US and screen for recent online shoppers of outdoor gear. I will also set a few values you can change: Desktop, Cam plus Screen plus Audio recording, and 10 participants. Want to keep those?

You: Make it 12 participants. Keep the rest.

Agent: Your study is ready to review. I’ve set everything up based on what you told me. Head to the Study Builder to make it yours. There you can review the screener logic, set up branching in your activities, add or remove activities, and fine-tune any detail before you launch.

Notice the good habits: a specific goal up front, a concrete task with a clear stopping point, a described audience, and a quick tweak (12 instead of 10) instead of re-explaining everything.

What the agent built

From that short exchange, here is the draft waiting in the Study Builder. Coherent, on-goal, and ready to fine-tune.

Study: Checkout drop-off, September. Unmoderated test of a live website, in English, targeting the United States. Desktop, Cam plus Screen plus Audio recording, 12 participants recruited from the Userlytics Panel.

Screener

S1. In the last 3 months, which of these have you shopped for online? (select all) Outdoor or camping gear (qualifies), Electronics, Groceries, Clothing, Home goods, None of these (disqualifies).

S2. How often do you shop online? Weekly (qualifies), A few times a month (qualifies), Monthly, A few times a year, Never (disqualifies).

S3. Which device will you use for this study? Desktop or laptop (qualifies), Mobile, Tablet.

Every question mixes real alternatives with the qualifying answer, so participants cannot guess their way in.

Activity script (11 steps)

1. Scenario. Welcome and instructions: you will shop on a real store and think out loud.
2. Task. Starting on trailhead-gear.com, find a rain jacket you like and add it to your cart.
3. Verbal. Now that you have tried that, anything stand out, good or bad?
4. Rating. How easy or difficult was adding to cart? 1 (very difficult) to 7 (very easy).
5. Verbal. Why that rating? What made it easy or hard?
6. Task. Continue to checkout, up to but not including the payment step. Stop there.
7. Verbal. Was there anything that made you hesitate, or that you did not expect?
8. Rating. How easy or difficult was the checkout? 1 to 7.
9. Verbal. Why that rating?
10. Single choice. Overall, how satisfied were you with the checkout experience?
11. Verbal. What is the main reason for your rating?

Each task is followed by an open probe and an ease rating, and the study stays on one goal: finding the friction in checkout.

If you remember five things

• Lead with your goal and the exact task you want people to try.
• Describe who you want to test with. That becomes your screener.
• Answer one question at a time, in plain words.
• Treat suggested values as a starting point and tweak what you want.
• Review and launch in the Study Builder. The chat only drafts.

Your next study is a conversation away. Open the AI Study Designer and start with one sentence about what you want to learn.

Userlytics, AI Study Designer. Example study shown for illustration. Trailhead is a fictional store.