Check on camera that participants show what you asked for
New for every organization. The AI research assistant checks the live camera for what you asked to see, asks again when an object is missing, and can screen out people who never show it.
Dag MalstafAt a glance
- Describe what the camera should show on any question in the study guide
- Participants hear exactly what to fix, from flipping the camera to the object still missing
- Several objects count even when they appear in different frames
- Setup checks flag requests that cannot be verified on camera before launch
- Filter interviews by the outcome of every visual check
Ask participants to hold up the coffee they drink and you often get half a second of it, out of focus, or the front camera still on their face. You find out in the recording, too late to ask again. Visual verification runs that check during the interview. From today it is available to every organization in new studies.
Short on time? Watch the one-minute version.
How it works
Say what the camera should show. In the study guide, turn on visual verification in a question's Live Vision options and describe what should be visible, for example "a mug or cup you usually drink from".
The AI research assistant checks the live camera. It uses the camera only: not the participant's screen, and not what they say or feel. It works the same on web, iOS and Android.
Participants see when it passes. A green "Visually Verified" badge appears once the camera has shown what you asked for.
It asks again when something is off. If an object is missing, unclear or barely on camera, the participant is asked to show it again, mid-conversation.
You decide what a failed check means. By default the interview carries on. Or end it on repeated failure, screening out anyone who has not shown the object after three attempts. The three-attempt limit is fixed.
What visual verification can do
Tell participants what to fix. Participants hear what is wrong: flip the camera, or "you showed the mug, please also show the coffee bag".
Check your setup before launch. The study guide flags requests that cannot be checked on camera or that contradict the question, and lists exactly what will be checked.
Detect several objects across frames. Each object counts once it has been on camera, even when the objects never share a frame.
Filter results by outcome. The interview detail page shows each check with its attempts and frames, and a dedicated facet filters interviews by the outcome.
Research teams asked for it in three situations
From teams in beauty, home care, food and drink, and product testing, kept anonymous.
Screeners that need proof. People who say they use a product show it, so a claimed user is a real one.
Home-use tests. Participants show the sample you shipped, so you know it arrived and is the product they are talking about.
The setup around the product. The washing machine and the detergents beside it, or the bathroom shelf. Each item counts once it has been on camera.
Ask for the product and keep faces out of frame.
What this means for your research
You get clear footage of what you asked about in every interview, not only when the participant got it right first time. Misunderstood tasks get fixed in the moment, and what the camera saw replaces a claim you had to trust.
Two limits. You can turn visual verification on in new studies only; a study that is already running keeps its current setup. And the check is only as good as what the camera sees: ask for things that can be shown, such as a product, a pack or a shelf, and treat a pass as strong evidence that still leaves room for error. For something out of reach during the interview, use picture upload.
Visual verification in one minute
A participant shows the product, the AI research assistant asks for what is missing, and the check passes on camera.
Watch Dag set it up and review the checks
Setup details and tips are in the visual verification guide.