Instagram has restricted the search terms, but its algorithm continues to display the videos regardless.

      The genre is straightforward to describe. A man wearing camera glasses approaches a woman on the street, delivers a line, and captures her reaction without permission. The video is uploaded to Instagram, tagged as shot on Meta glasses, and the woman appears satisfied. Many of these videos have amassed millions of views. Adam Mosseri, who oversees Instagram, stated in mid-July that this behavior needed to cease. Meta would eliminate “harassing” pickup videos per its existing bullying and harassment policies. We covered his commitment when he announced it. "We don't want individuals surreptitiously recording others and harassing them before posting on our platform," Mosseri mentioned at that time.

      In the month that followed, Business Insider investigated. Katie Notopoulos discovered numerous such videos still active, including several from popular verified accounts. She provided Meta with a list of over 20, of which the company removed nine. One notable removal involved a creator using the same line on multiple women in different videos. Meta deleted this instance, but the others remained available.

      Meta does acknowledge its actions. A spokesperson, Tracy Clayton, remarked that thousands of pieces of content have been taken down, and several large accounts have been removed. Instagram now restricts searches for terms like “rizz” and “cold approach”, which are slang associated with this type of content. These represent genuine enforcement actions, and no platform manages moderation flawlessly at such a scale.

      The ongoing pattern is the issue. One major account that was deactivated initially returned a few days later. Meta described this as a mistake and reinstated the ban after Business Insider brought it to attention. The following morning, Meta deactivated another account with 174,000 followers that appeared on the reporter’s list. After the owner appealed, it was reinstated the next day, with videos restored.

      This is not solely an Instagram issue. TikTok removed two out of four videos it received, citing bullying and sexual harassment. YouTube also took down several videos and suspended one channel from its monetization program.

      Now, the feed does the discovering

      A significant detail omitted from subsequent coverage is that after Notopoulos sought out and interacted with these videos, Reels began showing her more of them, until her feed was nearly filled with content of this genre. Search terms became unnecessary. Meta has blocked users from searching for these terms. However, the recommendation system still locates this content for viewers. Both systems, operated by the same company, are pulling in contradictory directions.

      This contradiction is more critical than any specific removal. Blocking a search term prevents intentional searching but does nothing about a ranking model that favors whatever captures attention, which evidently includes this type of content. Moderation eliminates items, while recommendations determine the visibility of those that remain.

      Researchers have begun to quantify the issue

      The concern is no longer just a journalistic one. Four researchers from the University of Sydney—Joanne Gray, Milica Stilinovic, Marcus Carter, and Ben Egliston—viewed hundreds of these videos and shared a preprint last week titled “Harm Through the ‘#RizzCam’: Smart Glasses, Ambient Capture and Invisible Harassment.” 404 Media reported on the paper on August 12. As it is a preprint, it has yet to undergo peer review.

      Carolina Are, a digital criminologist at the London School of Economics, told Business Insider that Meta's reaction overlooks the deeper level of the issue. “There is a backlash, and Meta tends to downplay its accountability. They refer to user-generated behavior or enforcement errors,” she stated. “However, they fail to acknowledge the systemic issues fueling this problem.”

      A problem created and marketed by Meta

      The reason Mosseri personally addressed this matter, unlike his counterparts at other platforms, is that Meta is involved on both fronts. The glasses are one of its products, and so is the feed. The harassment exists between these two elements.

      The hardware is achieving commercial success. Sales surpassed 7 million units in 2025, more than tripling from the previous year, and Counterpoint Research places Meta firmly in the lead regarding smart glasses. This achievement is a rare success for Reality Labs, which is still incurring losses and cut jobs this year. Mark Zuckerberg has argued that individuals without AI glasses will face cognitive disadvantages compared to those equipped with them.

      Reputation is at stake, and the backlash has grown over the past two months. A $2 sticker effectively concealed the recording light. The glasses were prohibited in UK pubs and subsequently in courts in England and Wales. Commenters labeled Kylie Jenner's posts, tagging them as showcasing Meta eyewear as “pervert glasses.”

      Automating the moderation process

      Meta's solution to managing moderation at scale is increased automation. In June, it informed the Financial Times that initial tests indicated its AI systems flagged more content violations than human moderators while making fewer errors. However, whether this holds true at scale remains unverified.

      Sarah T. Roberts, a co