Google recently published a blog post announcing updates to their machine learning systems designed to detect and remove more fake reviews, business listings, and fraudulent images and videos.
Their automated systems and human review teams removed over 200 million photos, 7 million videos, and blocked or removed over 115 million reviews, which is a 20% increase compared to 2021.
How Google Catches User-Contributed Spam
Google is now using new machine learning models to identify and eliminate fraudulent content. These models analyze unusual patterns in user-contributed content and identify new forms of abuse.
Google stated:
"We’ve long used machine intelligence to help us spot patterns of potential abuse, and we continue to evolve our technology.
Last year, we launched a significant update to our machine learning models that helped us identify novel abuse trends many times faster than in previous years.
For example, our automated systems detected a sudden uptick in Business Profiles with websites that ended in .design or .top — something that would be difficult to spot manually across millions of profiles.
Our team of analysts quickly confirmed that these websites were fake — and we were able to remove them and disable the associated accounts quickly."
Google’s systems review new content before it is posted to block fake or fraudulent submissions. They also use machine learning models to scan already published content to catch anything that may have slipped through.
These updated systems are faster and more effective than previous versions, catching more spam than in 2021.
Google explained:
"In some places, scammers started overlaying inaccurate phone numbers on top of contributed photos, hoping to trick unsuspecting victims into calling the fraudster instead of the actual business.
To combat this issue, we deployed a new machine learning model that could recognize numbers overlaid on contributed images by analyzing specific visual details and the layouts of photos.
With this model, we successfully detected and blocked the vast majority of these fraudulent and policy-violating images before they were published."
Spam Blocking Statistics
In 2022, Google:
- Blocked or removed over 115 million reviews, the majority of which were blocked before publication.
- Removed over 200 million photos and more than 7 million videos that violated their content policies.
- Blocked 20 million attempts to create fake business profiles.
- Added heightened protection for over 185,000 businesses experiencing suspicious activity.
In January 2023, Google communicated to the FTC their use of signals to identify fake accounts in addition to reviewing content. They now scan images to detect overlays designed to divert calls to scammers.
Various measures are used to identify fraudulent content, including checking for bots, duplicate content, and patterns similar to known fake reviews. They also utilize "intelligent text matching" to identify misleading content.
Authentic, Safe, and Reliable
Google employs both automated and human reviewers to prevent inauthentic activity on Google Maps. This effort is crucial for the users relying on business reviews and the businesses listed on the platform.
Featured image by Shutterstock/ViDI Studio
Source: Google
