In an ongoing effort to provide more transparency on how its local search and reviews engine works, Yelp today released its annual Trust & Safety report. It details how Yelp uses a combination of machine learning and human moderation to stop deceptive reviews, which is its own flavor of misinformation.

For example, Yelp shut down 278,600 user accounts in 2023 for various TOS violations. These accounts practiced myriad deceptive or abusive behaviors (detailed below) that engendered misinformation. It also removed 207,300 reviews containing inappropriate or harmful content such as hate speech.

In addition to large language models and lots of internal rigor, Yelp relies on its community of end users and SMBs to report bad actors and harmful content. Altogether, it’s about policing an influential source of content to maintain platform integrity and trust … as well as protect the good actors that occupy it.

“Fighting fake reviews and other deceptive behaviors are initiatives we’ve long invested in, built our platform around, and continue to evolve as we maintain a fair playing field for businesses that rightfully earn their great reputations,” Yelp VP of User Operations Noorie Malik told Localogy Insider.

Highlight Reel

What else did the 2023 Yelp Trust & Safety Report uncover? Here’s a highlight reel we’ve assembled to save you time…

• 207,300+ reviews were removed by Yelp’s content moderation team in 2023, including 28,800+ reported reviews removed for threats, lewdness, hate speech, or other potentially harmful content.
• 1M+ photos were removed for violating content policies.
• 278,600+ user accounts were closed for violating Terms of Service, including suspected deceptive or abusive behaviors.
• In 2023, Yelp integrated LLMs that detect and flag the most egregious instances of hate speech, vulgar language, or threats before they are published. This enabled the prevention of 23,600+ reviews
• 40,700+ potential new business pages were rejected for being associated with spammy behaviors in violation of Yelp’s policies — up 24% YoY.
• 1,500+ business pages were removed for being associated with lead generators – fake business pages to generate leads and auction them to other contractors.
• 53,000+ reported reviews were removed by Yelp’s moderators in 2023 for not reflecting firsthand customer experiences.
• 450+ warnings were issued to businesses after discovering someone may have engaged in compensated or incentivized review behaviors.
• 589 Suspicious Review Activity Alerts were applied after detecting a large number of suspicious positive reviews coming from a single IP address, or reviews from users who may be connected to a review exchange ring.
• 940+ user accounts were closed for being associated with review exchange rings or as part of Suspicious Review Activity Alert investigations.

Demonstrating Diligence

Stepping back, this report works on a few levels. The findings alone have value in terms of signaling the degree of attempted behaviors on Yelp, not to mention the level of policing that’s happening. But it’s also a signal to the world that Yelp is applying ample diligence to fighting misinformation on its platform.

That last part is key, as Yelp has been fighting PR battles since its inception in terms of maintaining its integrity as a trusted source of local business information. For example, there have been claims throughout the years that Yelp salespeople offer quid-pro-quo reviews for paid SMB accounts.

Altogether, such claims degrade Yelp’s standing as a source of local info. So it continues to fight those fires. Today’s report is one way to do that by shining a light on all of the diligence and investing it’s doing to keep the platform as a safe place, with local business information that can be trusted.

“These measures, among other efforts, help demonstrate how Yelp goes the extra mile to protect businesses, said Malik, “and aim to build trust with consumers by proactively addressing deceptive content and prevent bad actors from misleading our consumers.”