SAN FRANCISCO (Reuters) - Public databases that shine a light on online political ads - launched by Facebook and Google before Tuesday’s U.S. elections - offer the public the first broad view of how quickly the companies yank advertisements that break their rules.
The databases also provided campaigns unprecedented insight into opponents’ online marketing, enabling them to capitalize on weaknesses, political strategists told Reuters.
Facebook Inc and Google, owned by Alphabet Inc, introduced the databases this year to give details on some political ads bought on their services, a response to U.S. prosecutors’ allegations that Russian agents who deceptively interfered in the 2016 election purchased ads from the companies.
Russia denies the charges. American security experts said the Russians changed tactics this year. [nL4N1XH3JT]
Reuters found that 436 ads - 375 on Facebook and 61 on Google - from May through October related to 34 U.S. House of Representatives contests declared competitive last month by RealClearPolitics, which tracks political opinion polls.
Of the 258 removed ads with start and end dates, ads remained on Google an average of eight days and Facebook 15 days, according to data Reuters collected from the databases.
Based on ranges in the databases, the 436 ads were displayed up to 20.5 million times and cost up to $582,000, amounting to a fraction of the millions of dollars spent online in those races.
Asked for comment, Google said it is committed to bringing greater transparency to political ads. Facebook said the database is a way the company is held accountable, “even if it means our mistakes are on display.”
In some cases, the companies’ automated scans did not identify banned material such as hateful speech or images of poor quality before ads went live.
Ads that are OK when scanned may also become noncompliant if they link to a website that later breaks down.
Google’s database covers $54 million in spending by U.S. campaigns since May and Facebook $354 million, according to their databases.
Facebook’s figure is larger partly because its database includes ads not only from federal races but also for state contests, national issues and get-out-the-vote efforts.
The databases generally do not say why a particular ad was removed, and only Facebook shows copies of yanked ads.
The American Conservative Union political organization, which had 136 ads removed through Sunday on Facebook, said some commercials contained a brief shot of comedian Kathy Griffin holding a decapitated head meant to portray U.S. President Donald Trump.
Removing the bloody image resolved the violation for sensational content, and the organization said it had no qualms about Facebook’s screening.
Some removals were errors. The Environmental Defense Action Fund said Facebook’s automated review wrongly misclassified one of its ads as promoting tobacco.
Ryan Morgan, whose political consulting firm Veracity Media arranged attack ads for a U.S. House race in Iowa, said Google barred those mentioning “white supremacy” until his team could explain the ads advocated against the racist belief.
Five campaign strategists told Reuters they adjusted advertising tactics in recent weeks based on what the databases revealed about opponents’ spending on ads and which genders, age groups and states saw the messages.
Ohio digital consultant Kevin Bingle said his team reviewed opponents on Facebook’s database daily to take advantage of gaps in their strategy.
Morgan said his team tripled its online ad budget to $600,000 for a San Francisco affordable housing tax after Facebook’s database showed the other side’s ads were reaching non-Californians.
That political intelligence “let us know that digital was a place we could run up the score,” he said.
Reporting by Paresh Dave; editing by Greg Mitchell and Jonathan Oatis
Our Standards: The Thomson Reuters Trust Principles.