Watch an advertisement on a video screen in a mall, health club or grocery store and there's a slim — but growing — chance the ad is watching you too.
Small cameras can now be embedded in the screen or hidden around it, tracking who looks at the screen and for how long.
The makers of the tracking systems say the software can determine the viewer's gender, approximate age range and, in some cases, ethnicity — and can change the ads accordingly.
That could mean razor ads for men, cosmetics ads for women and video-game ads for teens.
And even if the ads don't shift based on which people are watching, the technology's ability to determine the viewers' demographics is golden for advertisers who want to know how effectively they're reaching their target audience.
While the technology remains in limited use for now, advertising industry analysts say it is finally beginning to live up to its promise.
The manufacturers say their systems can accurately determine gender 85 to 90 percent of the time, while accuracy for the other measures continues to be refined.
The concept is reminiscent of the science-fiction movie "Minority Report," in which Tom Cruise's character enters a mall and finds that retinal scanners identify him and prompt personalized ads that greet him by name.
But this technology doesn't go nearly that far. It doesn't identify people individually — it simply categorizes them by outward appearances.
So a video screen might show a motorcycle ad for a group of men, but switch to a minivan ad when women and children join them, said Vicki Rabenou, the chief measurement officer of Tampa, Fla.-based TruMedia Technologies Inc., one of the leaders in developing the technology.
"This is proactive merchandising," Rabenou said. "You're targeting people with smart ads."
Because the tracking industry is still in its infancy, there isn't yet consensus on how to refer to the technology. Some call it face reading, face counting, gaze tracking or, more generally, face-based audience measurement.
Whatever it's called, advertisers are finally ready to try it, said advertising consultant Jack Sullivan, a senior vice president of Starcom USA in Chicago.
"I think you're going to see a lot of movement toward it by the end of this year in the top 10 markets," he said.
Because face tracking might feel reminiscent of Big Brother, manufacturers are racing to offer reassurances.
When the systems capture an image of who's watching the screen, a computer instantly analyzes it.
The systems' manufacturers insist, however, that nothing is ever stored and no identifying information is ever associated with the pictures.
That makes the system less intrusive than a surveillance camera that records what it sees, the developers say.
The idea still worries Lee Tien, a senior staff attorney with the Electronic Frontier Foundation, a civil-liberties group in San Francisco.
Tien said it's not enough to say some system is "not as bad as some other technology," and argues that cameras that study people contribute to an erosion of privacy.
In general, the tracking systems work like this: A sensor or camera in or near the screen identifies viewers' faces by picking up shapes, colors and the relative speed of movement.
The concept is similar to the way consumer cameras now can automatically make sure faces are in focus.
When the ad system pinpoints a face, it compares shapes and patterns to faces that are already identified in a database as male or female. That lets the system predict the person's gender almost immediately.
"The most important features seem to be cheekbones, fullness of lips and the gap between the eyebrows," said Paolo Prandoni, chief scientific officer of Quividi, a French company that is another player in face-tracking technology.
Others include Studio IMC Inc. in New York.
The companies say their systems have become adept at determining a viewer's gender, but age is trickier: The software can categorize age only in broad ranges — teens, younger to middle-aged folks and seniors.
There's moderate demand for ads based on ethnic information, but the companies acknowledge that determining ethnicity is more challenging than figuring out gender and age range.
Prandoni provided The Associated Press a limited version of Quividi's software, which uses an ordinary webcam to stream video to a computer. The trial version tracked gender only, using color-coded circles to distinguish male and female faces.
The sample size was too small to be statistically significant, but it was accurate about 80 to 90 percent of the time.
That might be as precise as the systems ever get, said Deborah Mitchell, a professor of consumer psychology at the University of Wisconsin-Madison. Even the human brain can't always determine gender, age or ethnicity.
Still, "even if it gets to 70 percent accuracy, that's still giving you a wealth of information," said Mitchell, who teaches in the Wisconsin School of Business.
That information is certainly valuable to Bill Ketcham, the chief marketing officer of Adspace Networks Inc. His New York company sells video advertising on 1,400 video screens at 105 malls around the nation.
Adspace is testing six TruMedia systems at malls in Winston-Salem, N.C., Pittsburgh and St. Louis. The kiosks display a daily list of top 10 sales at the mall, as well as paid advertising that comes largely from movie studios and TV networks.
A 15-second video ad that replays across Adspace's national network can cost as much as $765,000 per month. So advertisers expect rigorous information about who sees the spots — information that face tracking can now provide, Ketcham said.
For now, at least, Adspace isn't changing the ads based on who's watching — Ketcham said the kiosks' audiences are so large that it wouldn't be practical to personalize ads to individuals.
While advertisers like the face-tracking technology, another privacy advocate, Harley Geiger, questions whether it should be used on consumers without their knowledge.
Geiger, staff counsel for the Center for Democracy & Technology in Washington, D.C., said advertisers should be telling consumers what details about them are being collected and for what purpose.
"With the technology proliferating, now or the short-term is the time to consider privacy protections," he said. "If you don't build it in at an early stage it becomes very difficult to build it into an already established system."