If your Browse.ai robot misses items on a paginated or infinite-scroll page, retrain the navigation action, not the whole robot. The robot learns field structure from the categories you assign (title, price, colors) plus the interaction flow - scrolling and next-page clicks are trainable steps. Most 'it only got the first page' complaints are a missing next-page action in the training, not a platform limit: Browse.ai handles pagination, infinite scroll, and JS pages when those steps are part of the trained flow.

Context: Browse.ai guides: robots support dynamic website features - pagination, infinite scroll, and JavaScript-rendered pages - as part of the trained flow. The training model: when training a list capture, you assign categories to each text field (title, pricing, colors) so the robot learns the field structure, not just the selectors. If extraction quality is poor on a dynamic page, the fix is usually re-training the specific action (e.g. the scroll or next-page click) rather than rebuilding the whole robot.