Developed by scientists at the Massachusetts Institute of Technology, the system looks at how muscles in the face move in response to watching a video, then software classifies what counts as positive facial responses and smiles during the video.
From that it can be predicted which advertisement the viewer most enjoyed, NewScientist.com reported Wednesday.
The researchers collected more than 3,200 videos of study participants whose faces were filmed by their own computer's webcam as they watched three advertisements online during the Super Bowl in 2011.
The smile intensity was tracked during the video and then used to predict whether they would respond positively or negatively when subsequently asked if they enjoyed the ad and whether they would want to see it again.
In tests, the system made correct predictions more than 75 percent of the time.
The system could be used to tailor advertisements for viewers as they watch programs online based on their reaction, or as a more effective way of testing how appealing an ad is to customers, the researchers said.