Concept and creative testing with AI interviews
By Juno · Published · Updated
Concept and creative testing with AI interviews means showing real people several routes or concepts and following up on each reaction until the reason behind it is clear. Done well, it tells you which route people prefer and why, including where the audience splits, which a score on its own tends to hide.
What interviews add to a concept test
A rating tells you how a route scored. It rarely tells you what to change, or who the route worked for. In an interview, a person can say what they took the route to mean and whether it felt like it was for them, and the interviewer can ask why.
AI-moderated interviews make it practical to do this with many people rather than a handful. Each person has a real conversation, by text or voice, in their own browser. Open conversation can sit alongside structured questions, such as a rating or a ranking, in the same interview, so you get a comparable measure as well as the reasons behind it.
Preparing the stimuli
Juno can show images or external links during an interview. A route might be a headline with a key visual, a moodboard, a claim, a product page or a prototype. A few practical points:
-
Make routes comparable. If one route is a finished visual and another is a rough sketch, people react to the finish rather than the idea. Bring each route to a similar level of polish, or tell participants they are early ideas.
-
Label routes neutrally. Use Route A and Route B, not working titles such as "Bold" or "Safe" that hint at your thinking.
-
Keep each stimulus focused. One idea per stimulus is easier to react to than a board that mixes several.
-
Decide what must be asked word-for-word. If you need the same question for every route, such as a rating, mark it as verbatim. Juno asks it as written and adapts the follow-up questions around it.
Keeping the interview from leading towards a favourite
Most teams have a favourite route, and it can leak into the study. The common ways it happens:
-
Order. The first route people see sets the reference point for the rest. If your favourite always comes first, or always comes last, it is hard to tell whether people prefer the idea or its position.
-
Wording. "What do you like about this one?" assumes they like it. "What does this say to you?" doesn't. Ask the same neutral question of every route.
-
Context last. Showing the work before you understand the person means you can't tell whether a route fits how they think about the category or is simply new. Start with their own situation, in their own words, then show the routes.
-
Uneven probing. If the interviewer digs into objections to one route and accepts praise for another at face value, the comparison is skewed. Every route needs the same depth of follow-up.
Juno checks a study's design for loaded wording and built-in assumptions before anyone is interviewed, and pushes back on a brief that leans towards one answer. Interviews are led by research objectives, so Juno can ask openly first and then follow what each person says. For more on wording, see leading questions in research.
Comparing routes
Look at who preferred each route and why, not only at which route came out on top. Two things are worth looking for:
-
Splits. A route can be the clear favourite of one group and the first to be rejected by another. An average hides that, and it is often the most useful finding in the study.
-
Reasons that travel. If people reject every route for the same reason, such as confusion about what the product does, that points to the proposition rather than the creative.
It helps to ask for a choice as well as reactions. A ranking or a single-choice question near the end of the interview gives you a clear preference to set against the reasons people gave.
Juno's synthesis interprets what people said rather than counting how often words come up. Each finding is stated as a claim with the evidence behind it, and confidence wording is matched to the strength of that evidence. In the debrief, you can ask where the audience splits on a route and pull up the exact moments behind the answer. Which route to take forward stays your decision.
Example: Amber Electric
Amber Electric, an Australian battery plan, interviewed 100 battery owners with Juno: some with a battery already on the wall and some about to get one. Juno started with the owners themselves, including why they bought and how they use the battery day to day. Once it understood how each owner thought about their battery, it moved on to a few of Amber's positioning ideas and asked what they made of each one.
One message divided the room. Owners who liked it chose it more often than any other option, and owners who didn't rejected it more often than any other. In a survey, those two groups would have cancelled each other out, and the message would have scored somewhere in the middle. In the interviews, owners explained in their own words what they liked or disliked about it, which gave Amber a clear sense of how to use the message and who to use it with.
Amber left knowing which messages to use and which to rework. "It goes beyond what you'd expect by nudging participants and digging in with nuance," said Mick Oechsle from Amber's design team. Read the Amber Electric story.
How Juno handles this
Brief Juno on the routes you are testing and the decision they feed. Juno builds the study around research objectives and checks the design for leading wording before anyone is interviewed. In each interview, Juno shows the routes and follows the reasons behind each reaction, while keeping word-for-word questions intact so the routes can be compared across the study. Juno is built on 25+ years of qualitative research experience. See how Juno interviews.
Frequently asked questions
What can I show participants in an AI interview? Images or external links. That covers most early creative, such as a key visual, a claim, a product page or a prototype.
How do you stop people being polite about the work? Ask what each route says to them before asking whether they like it, and ask for concrete reasons. Juno asks for a concrete example rather than accepting a polite abstraction, and nudges participants whose answers are too brief.
Can I combine ratings with open conversation? Yes. Rating scales and ranking questions can sit in the same interview as open conversation, so you get a clear preference and the reasons behind it.
Does AI concept testing use synthetic respondents? Not with Juno. Every finding comes from real people. Simulated interviews are used only to rehearse a study before launch, never as research data.