Yelp’s annual Trust & Safety report is out today. Every year, it reveals how Yelp’s local search and reviews engine works, including reliance on machine learning and human moderation to police deceptive reviews. As part of this exercise, it also quantifies its policy enforcement efforts around bad actors.
For example, Yelp shut down 551,200+ user accounts in 2024 for TOS violations. These accounts practiced a range of deceptive or abusive behaviors that mostly boil down to misinformation. It also removed 185,000+ reported reviews containing harmful content such as hate speech.
This year, the report has an additional layer of relevance given recent inflections in AI. Here, Yelp is careful to distinguish good and bad uses of AI. For example, it uses a mix of AI, human moderation, and community feedback to police review integrity. But it disparages AI’s use to write reviews.
“At Yelp, we require all reviews to be based on genuine firsthand consumer experiences. AI shouldn’t be a shortcut for writing reviews, Yelp SVP of Consumer Product, Akhil Kuduvalli told Localogy Insider. “By leveraging AI to help summarize or better surface our rich content, we’re enabling consumers to more seamlessly discover great local businesses. All of this is built on top of and powered by authentic experiences shared firsthand by our community, not an AI application.”
Highlight Reel
What else did the 2024 Yelp Trust & Safety Report uncover? Here’s a highlight reel we’ve assembled to save you time…
- 185,100+ reported reviews were removed by Yelp’s User Operations team in 2024.
- 551,200+ user accounts were closed in 2024 due to Terms of Service violations, resulting in the removal of more than 415,300+ reviews.
- 39,200+ potential new business pages were rejected in 2024 for being associated with spammy behaviors, violating Yelp’s policies.
- 920 Media Attention Alerts were placed on business pages in 2024 related to a business gaining public attention, resulting in the removal of 56,900+ reviews.
- 1,300+ user accounts were closed for being associated with review exchange rings or other suspicious review
activity in 2024. - 920+ reports were made by Yelp to other online platforms in 2024, warning them of suspicious groups, posts, or individuals on their sites (e.g., Instagram, Facebook, LinkedIn, and Reddit). 37% of those reports resulted in action being taken by these online platforms to remove deceptive
content. - 554 business pages received an Unusual Activity Alert or Public Attention Alert in 2024, resulting in the removal of 41,500+ reviews, for incidents stemming from social media (TikTok, Facebook, X, etc.).
- 188 business pages received an Unusual Activity Alert or Public Attention Alert in 2024, resulting in the removal of 20,700+ reviews, for incidents related to alleged discrimination (including racism, antisemitism, anti-LGBTQ+ sentiment, ableism, etc.).
- 219 business pages received an Unusual Activity Alert in 2024, resulting in the removal of 14,700+ reviews, for incidents related to politics or political figures.
- 155 business pages received an Unusual Activity Alert in 2024, resulting in the removal of 5,600+
reviews, for incidents related to alleged criminal implications or activity (including claims of animal
cruelty, homicide, assault, child neglect, etc.).
White Hat AI
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, given that reviews can be inherently triggering for business owners, they’ve become a lightning rod for contentious energy.
That backlash can challenge Yelp’s standing as a source of local info. So it continues to defend its credibility. Today’s report is one way to do that by shining a light on its diligence in keeping the platform pure. As noted, it’s using AI to help achieve that end, rather than degrade truth. We’ll call it white-hat AI.
“This reinforces our commitment to authentic content and shapes the features we develop,” said Kuduvalli. “As an example, AI can add value to the review ecosystem by summarizing and highlighting unique experiences from genuine reviews, helping users quickly find the right business. Our AI-powered features, like Review Insights and review highlights, use LLMs to emphasize key sentiments for common topics and surface unique aspects of a business from reviews based on the consumer’s search intent.”
