Number of bad accounts removed using machine learning has doubled to more than a million.
Google took down 2.3 billion “bad ads” last year, which is significantly down on the 3.2 billion it removed the year before.
The internet’s dominant ad company has also provided more detail about the kinds of ads it removed last year for violating its policies.
These include: 207,000 ads for ticket resellers, more than 531,000 ads for bail bonds and about 58.8 million phishing ads, while it removed nearly 734,000 publishers and app developers from its ad network.
However, Google could not provide an estimate for how big of a proportion 2.3 billion is against its total ad inventory that was served in 2018, although it is understood to be a very small minority.
Nor would it comment on how much income it made from those ads.
Google has been at the centre of multiple brand safety scandals in recent years, such as ads appearing next to terror videos and extremist content.
However, last year Google was able to identify and remove almost one million bad advertiser accounts using machine learning, which is nearly the double amount it terminated.
Because of the sheer volume of new web pages that are created every day, as well as videos posted on Google’s video-sharing platform YouTube, the company has tried to develop artificial intelligence-led solutions to moderation instead of just relying on humans.
Scott Spencer, director of sustainable ads at Google, said: “Google has a crucial stake in a healthy and sustainable digital advertising ecosystem – something we’ve worked to enable for nearly 20 years.
“Every day, we invest significant team hours and technological resources in protecting the users, advertisers and publishers that make the internet so useful.”