How machine learning keeps contributed content helpful

Contributed content about local businesses plays a big part in helping people find the latest information — whether they’re looking for a grocery store that’s open late on Maps or seeing what people have to say about a new neighborhood bakery on Search. In 2023, we received around 20 million contributions per day on Maps and Search, which included updates to business hours, ratings, photos, reviews, videos and more. To make sure that this information remains accurate in a rapidly changing world, we use a combination of technology and skilled operators, and collaborate regularly with the wider industry to fight bad actors.

Here’s a look at how we’ve updated our systems to help us identify and take down more fake reviews than ever before, and how we used it in combination with other techniques to stop a widespread internet scam. Plus, we’re sharing the progress we made in 2023 to keep contributed content helpful for everyone.

A new machine learning algorithm helps us catch fake reviews faster

Reviews undergo rigorous scrutiny by our moderation systems before they’re published, but the work doesn’t stop there. Our systems continuously monitor contributed content and watch for suspicious activity over time. Last year, we launched a new machine learning algorithm that detects questionable review patterns even faster. It does so by examining longer-term signals on a daily basis — like if a reviewer leaves the same review on multiple businesses or if a business receives a sudden spike in 1 or 5-star reviews. In 2023, this new algorithm helped us take down 45% more fake reviews than the year before.

How the new algorithm in action helped stop scammers in their tracks

This update works to catch both one-off cases and broader attack patterns. For example, a network of scammers falsely claimed that for a low fee they would connect people to high-paying online tasks, like writing fake reviews or clicking ads across the internet. Our new algorithm quickly identified this surge in suspicious reviews thanks to its ability to continuously analyze patterns, like whether an account had previously posted reviews on other targeted features.

From there, our team of investigators analyzed reports from merchants who recently saw a spike in fake 5-star reviews, and even worked with people who were contacted by scammers on other platforms. Once we pinpointed patterns, like what categories of businesses were being targeted, we refined our algorithms to help protect hundreds of targeted businesses from the bad actors.

In just a few weeks, we caught over five million fake review attempts related to this scam. To prevent this scam from impacting others, we quickly disabled the malicious accounts associated with the scam and placed protections on hundreds of businesses. We also shared tips with users about spotting scams so they could stay informed and vigilant. Thanks to our rapid detection and action, we were able to address this issue and protect our community.

Our progress in 2023

In 2023, we made significant strides against fake content thanks to our updated technology and skilled analysts. Here are some of the key actions we took last year that protected local business reviews on Google Maps and Search.

• Thanks to our evolving machine learning algorithms, we blocked or removed over 170 million policy-violating reviews from 2023 — over 45% more than in 2022. More than 12 million fake business profiles were removed or blocked too.
• Improvements to video moderation algorithms, like detecting fake overlaid phone numbers, helped us catch 14 million policy-violating videos in 2023 — 7 million more than the previous year.
• Business owners were protected from over 2 million attempts by bad actors trying to claim Business Profiles that did not belong to them — over 1 million more than in 2022. And temporary protections were placed on more than 123,000 businesses after our systems detected suspicious activity and abuse attempts.

Legal action and broader industry collaboration play an important part in our approach to combatting deceptive content too.

• In 2023, we filed a lawsuit against a malicious actor who had been posting fake reviews on Google Maps and fraudulently attempting to manipulate services for small businesses.
• Some examples of our ongoing industry collaboration include sharing expert insights during the FTC’s advanced notice period for combating deceptive content on the internet, and participating in the proposed rulemaking discussions as a member of the Computer & Communications Industry Association.

Reliable information is vital to helping people explore the world, and we know that deceptive content erodes that trust. We’re committed to combating fake reviews, while continuing to collaborate with regulators and the broader industry on this important issue.