The research could influence public health policy in countries where the technique is used.
Electronic dolls designed to reduce the rate of teenage pregnancy do not work, a new study has found, and may even increase the number of babies born to young mums.
Researchers found that girls between the ages of 13 and 15 who were given a simulator infant to look after were actually more likely to become pregnant early in life than those who had simple sex education.
The research could influence public health policy in countries where the technique is used, including Australia, the United States and Britain.
Around 3,000 youngsters took part in the study in Western Australia over three years from 2003, with researchers tracking their medical records until they reached the age of 20.
Half were given a life-like electronic doll, which needed to be cared for and cried like a real baby, for a weekend.
The other half, the control group, received only standard health education, according to the study published in The Lancet Thursday.
Lead researcher Sally Brinkman said the group given the robot baby were more likely to become pregnant before early in life.
Of those who had charge of a doll, 17 percent recorded at least one pregnancy -- whether carried to full term or terminated -- by the age of 20.
The rate in the control group was just 11 percent, which researchers described as a small but statistically significant difference.
"Unfortunately that's the finding," Brinkman told AFP.
Brinkman said researchers at Western Australia's Telethon Kids Institute were trying to determine whether the robot babies, which she said cost over US$1,000 each, were worth the money.
She said while some of those researching the life-like dolls thought they were perhaps ineffective, no one had expected them actually in increase the likelihood of an early pregnancy.
"We never went into the study thinking this would increase teen pregnancy," she said.
Brinkman said the research did not reveal why the girls given the toy babies were more likely to fall pregnant at a young age.
"But from delivering the programme to that many girls, we got feedback," she said.
"And in general most of the students really liked it (the programme). You do get a lot of attention when you have (a pretend baby)."
Of all the girls who fell pregnant, 53.8 percent of those who had the robot baby terminated the pregnancy compared with 60.1 percent in the control group.
The researchers said while that difference was not huge, it indicated participants who had exposure to the robot baby appeared more likely go through with the pregnancy.
Brinkman said overall she could not recommend the idea of simulator infants if educators wanted to prevent teen pregnancy.
"The Virtual Infant Parenting Programme is used across Australia and the world because it is thought to reduce rates of teen pregnancy," she said in a statement.
"This is the largest study of its kind and highlights that even the most well-intentioned programmes can have unexpected consequences."