From Scroll to Stop: A Real Test
Video · 2026-07-22
Watch a real user test and see how we analyze social media content with Stopper AI.
What you will watch
A real test from start to finish: a set of social media visuals goes through Stopper Analyze with real participants, then through Stopper AI, and the two results are compared. No staging; the feed, the clicks and the heatmaps are recorded as they happened.
Step by step
Upload. The visuals are added in the order they will appear. Each one is cropped to the platform frame.
Settings. Eye tracking on, reactions on, shuffled order, distractor posts mixed in, a time limit for equal exposure.
Share. One link. Participants open it on their own phone, enter a name and e-mail, calibrate the webcam and scroll.
The feed. Nothing is highlighted. The test visuals sit between realistic filler posts, and participants scroll the way they always do.
The report. Image ranking, stopping rate, dwell, first-view timing, clicks and reactions per visual, gaze and cursor heatmaps, and a one-paragraph summary of what happened.
Then the prediction
The same visuals are uploaded to Stopper AI. Nine analyses run: predicted fixations, heatmap intensity, saliency, composition, typography, colour and contrast, negative space, an EEG-style simulation and the aggregation layer. Each visual gets a score and written recommendations.
What to compare
Where the prediction and the behaviour agree, the score can be trusted for that kind of visual. Where they disagree, the Lab data is the teacher: those cases feed back into calibration. The video ends on exactly that comparison for the test set.
Try it
Any account can run the same flow in Stopper Analyze. The analyses in the video are the standard ones; nothing was tuned for the recording.