Google is a tech powerhouse in many categories, including advertising. Today, as part of its efforts to improve how that ad business works, it provided an annual update that details the progress it’s made to shut down some of the more nefarious aspects of it. Using both manual reviews and machine learning, in 2018, Google said removed 2.3 billion “bad ads” that it said violated its policies, which most generally forbid ads that mislead or exploit vulnerable people. Along with that, Google has increasingly started to tackle the “bad ads” conundrum on the other side: pinpointing and shutting down sites that violate policies that are profiting from using its ad network: it said it removed 1.5 million apps and nearly 28 million pages that violated publisher policies.
On the more proactive side, the company said today that it is introducing a new Ad Policy Manager in April to give tips to publishers to avoid listing non-compliant ads in the first place.
Google’s ad machine makes billions for the company — more than $32 billion in the previous quarter, accounting for 83 percent of all Google’s revenues. Those revenues underpin a variety of wildly popular, free services such as Gmail, YouTube, Android and of course its search engine — but there is undoubtedly a dark side, too: bad ads that slip past the algorithms and mislead or exploit vulnerable people, and sites that exploit Google’s ad network by using it to fund the spread of misleading information.
Notably, Google’s 2.3 billion figure is nearly 1 billion less ads than it removed last year for policy violations. The lower numbers might be attributed to two things. First, while the ad business continues to grow, that growth has been slowing just a little in competition with other players like Facebook and Amazon. Second — and this one gives the benefit of the doubt to Google — you could argue that it has improved its ability to track and stop these ads before they make their way to its network. (The more cynical question is whether Google removed less to improve its bottom line.)
Google’s director of sustainable ads, Scott Spencer, highlighted ads removed from several specific categories this year: there were nearly 207,000 ads for ticket resellers, 531,000 ads for bail bonds and 58.8 million phishing ads taken out of the network.
Part of this was based on the company identifying and going after some of these areas, either on its own steam or because of public pressure. In one case, for ads for drug rehab clinics, the company removed all ads for these after an expose, before reintroducing them again a year later. Some 31 new policies were added in the last year to cover more categories of suspicious ads, Spencer said. One of these included cryptocurrencies: it will be interesting to see how and if this one becomes a more prominent part of the mix in the years ahead.
Because ads are like the proverbial trees falling in the forest — you have to be there to hear the sound — Google is also continuing its efforts to identify bad apps and sites that are hosting ads from its network (both the good and bad).
On the website front, it also created 330 “detection classifiers” to seek out specific pages that are violating policies. Google’s focus on page granularity is part of a bigger effort it has made to add more granular tools overall to its network — it also introduced page-level “auto-ads” last year — so this is about better housekeeping as it works on ways to expand its advertising business. The efforts to use this to ID “badness” at page level led Google to shutting down 734,000 publishers and app developers, removing ads from 1.5 million apps and 28 million pages that violated policies.
Fake news also continues to get a name check in Google’s efforts. The focus for both Google and Facebook in the last year has been around how its networks are used to manipulate democratic processes. No surprise there: this is an area where they have been heavily scrutinised by governments. The risk is that, if they do not demonstrate that they are not lazily allowing dodgy political ads on their network — because after all they still represent ad revenues — they might find themselves in regulatory hot water, with more policies being enforced from the outside to curb their abuses.
This past year, Google said that it verified 143,000 election ads in the US — it didn’t note how many it banned — and started to provide new data to people about who is really behind these ads. The same will be launched in the EU and India this year ahead of elections in those regions.
The new policies it’s introducing to improve the range of sites it indexes and helps people find are also taking shape. Some 1.2 million pages, 22,000 apps and 15,000 sites were removed from its ad network for violating policies as misrepresentative, hateful or other low-quality content. These included 74,000 pages and 190,000 ads that violated its “dangerous or derogatory” content policy.