AI Search Practical insights
Test product discovery with Search Live
Design a practical Search Live study to learn what shoppers ask when they point a camera at a product and where your product information falls short.
A visual product discovery study should reveal what a shopper needs to understand, not merely whether an AI system recognizes an object. A correctly identified lamp can still leave unanswered questions about its fitting, dimensions or suitability for a room. Those gaps are actionable marketing research.
Observe a real selection task
Choose a small set of products with meaningful differences. Ask participants to find a suitable desk lamp for a cramped workspace, for example. Give them access to the visual search experience available on their own device, and record whether Search Live is actually accessible in their country and language. Do not describe a feature as universally available because it appeared in a launch presentation.
Separate identification from suitability
Record the object shown to the camera, the participant’s words and any extra context they provide. Then distinguish identification errors from unanswered buying questions. A wrong model number is an identification problem. An accurate model with no information about glare is an evidence problem. The corrective action differs: one may require clearer imagery and identifiers; the other may need a useful explanation or demonstration.
Use a controlled comparison
Repeat the task with a conventional text search, changing the order across participants so familiarity does not always favor the second method. Compare time to a defensible choice, unresolved constraints and confidence supported by evidence. A faster answer is not necessarily a better purchase decision. Keep observations descriptive when the sample is small.
Improve the product page that carries the answer
If several participants struggle with the same constraint, add accurate dimensions, compatibility details or contextual photography to the relevant product page. In the lamp example, a photograph beside a monitor can help communicate scale, but it does not replace numerical dimensions. Re-run the task after the update and look for fewer unresolved questions. The test evaluates buyer understanding; it does not prove that a page change causes AI visibility to rise.
Sources and evidence
Sources checked on 4 October 2026. Proposed workflows and hypothetical examples are editorial analysis.