Skip to main content
Visual verification lets the AI moderator attempt to check what a participant is showing on camera before continuing. Use it when an answer is only meaningful if the participant is actually holding, using, or showing the thing you asked about — a product, a packaging, a label, a setting on their phone. While the question is active, the moderator samples frames from the participant’s webcam and checks them against a short description you write. The participant sees a small badge confirming they’ve been verified, but this does not guarantee that every condition of the task was satisfied.
Visual verification is currently available to selected organizations. If you don’t see the Visual verification option in the question editor, it isn’t enabled for your workspace yet — reach out to your Conveo contact.

​
When to use it

Visual verification is for open-ended questions where you want a visual check that the participant is engaging with something physical or on-screen:
  • “Show us your coffee machine and tell us about it.”
  • “Hold up the packaging — what stands out?”
  • “Open the app’s settings screen and walk us through what you see.”
It is not a screener and not a quality check. It runs on a single question at a time, and only after the participant has already joined the interview.

​
Turning it on for a question

  1. Open your Topic Guide.
  2. Click an open-ended question to edit it.
  3. Open the Live vision dropdown (under the question text).
  4. Choose Visual verification.
A small panel appears under the question with a What to verify field.

​
Writing the verification description

Write a short, plain-language description of what should be visible. The moderator turns this into the instructions the AI uses on each frame. Good descriptions are concrete and visual:
  • “show their coffee machine”
  • “hold up the back of the package”
  • “point the camera at the dog’s food bowl”
Avoid descriptions that require reading text in fine print, judging brand authenticity, or any check that a human would struggle to make from a single video frame. After you save the description, the status pill in the panel goes from Generating prompt… to Active. Once it says Active, the question is ready to run. If generation fails, the pill turns red with a Regenerate prompts button — click it to retry. Editing the description always regenerates automatically.

​
What the participant sees

Nothing changes about how the question is asked. The moderator reads the question, the participant answers, and frames are analysed in the background. When the check passes, a green Visually Verified badge appears for a few seconds, then fades away. A passed visual check does not bypass the question’s other response handling. If the check returns a definitive failure, the moderator can ask the participant to adjust — for example, “Could you point the camera a bit lower so I can see the machine?” — and retry within the retry limit. A missing or undetermined result does not trigger the same retry sequence. By default, exhausting the visual retry budget does not end the interview solely because verification failed.

​
End interview on repeated failure

By default, visual verification is soft: verification failure alone does not end the interview after the retry budget is exhausted. If a repeated, definitive visual failure should end the interview, enable End interview on repeated failure in the visual verification panel:
  1. Open the visual verification configuration on the question.
  2. Enable End interview on repeated failure.
The runtime allows two retries after the initial attempt, then ends the interview if the last verdict is a definitive failure and hard-cancel mode is enabled. An undetermined verdict, missing result, or insufficient visual evidence is not treated as a definitive failure. Hard-cancel mode therefore does not guarantee that every continuing participant was successfully verified.

​
Tips

  • One thing per question. Verifying “the coffee machine and the milk frother” is harder than verifying just the machine. Split it across two questions if you need both.
  • Match the camera reality. Most participants are on a laptop webcam or phone. A description like “show the underside of the box” works; “show the serial number” usually doesn’t — the camera can’t focus that close.
  • Test it. Run the study on yourself before launch (see Testing your study) and walk through the verification flow as a participant would. If you can’t pass your own check easily, the description needs tightening.
  • Check failure behavior. Decide whether a definitive visual failure should end the interview, and test the setting. A missing camera signal is not proof of either success or failure.
  • Treat detection as fallible. Sampled frames can miss brief actions or misidentify an object. Visual verification is not a guarantee of product authenticity, participant identity, or compliance.