Google, on Wednesday disclosed data showing that while its pay-per-click Web advertising system is under regular attack from fraudsters, virtually all such tricks are automatically detected, rebuffing critics who say its online ads are a magnet for fraud.

Click fraud can occur when Web site publishers attempt to trick Google’s ad system into counting ads never seen by real users, or when competitors use automated programs to create fake clicks, driving up charges per click and rivals’ ad rates.

Invalid clicks also regularly occur if consumers click on an ad then instantly click the back button to exit the ad. Google discounts both.

Financially, what this means for Google is that its computers automatically reject up to 10 percent of potential advertising billings. At current revenue levels, every percentage point of invalid clicks the company forgoes costs Google $100 million in lost revenue, the product manager said.

Shuman Ghosemajumder, Google’s product manager for trust and safety, said in an interview that, on average, up to 10 percent of pay-per-click activity is invalid, and in some cases fraudulent, but that its computers automatically detect virtually all such problems, meaning advertisers pay nothing.