X expels 4,000 creators from revenue program using Grok to detect plagiarism
Elon Musk's social network deploys an enhanced version of its AI that triples the speed of duplicate content detection, removing thousands of accounts that reposted others' content without attribution.
July 20, 2026 · 5 min read
TL;DR: X has expelled over 4,000 creators from its revenue sharing program after detecting via Grok that they reposted others' content. The AI now identifies duplicates three times faster, protecting original creators and redefining incentives on the platform.
What happened?
X, the social network owned by Elon Musk, has announced the expulsion of over 4,000 creators from its revenue sharing program after detecting they reposted other users' content without attribution. The detection was carried out thanks to an update to its Grok AI model, which now identifies duplicate content (videos, images, and text) at a speed three times faster than the previous version. The information was confirmed by Nikita Bier, head of product at X, and reported by TechCrunch on July 16, 2026. According to Bier, the new version of Grok can process and compare metadata, video fingerprints, and text patterns much more efficiently, allowing it to detect violations before they accumulate large volumes of engagement. This announcement comes in a context where X seeks to strengthen its value proposition for original creators, following the decline in advertising revenue and the exodus of advertisers in 2023-2024.
The structural problem: the copy economy
X's revenue sharing program rewards creators for engagement generated among X Premium subscribers. This created a perverse incentive: download viral videos from other platforms or smaller accounts, repost them as one's own, and reap the engagement and money that would belong to the original creator. Bier had already warned in May 2026 about the existence of "a significant number of large accounts" that engaged in this practice programmatically. He cited the case of an account with 1.6 million followers that used a secondary account to watermark another creator's video with its own brand and then reposted it on the main account. This phenomenon is not new in social media: platforms like YouTube have dealt with 'freebooting' for years, implementing systems like Content ID in 2007. However, X's scale and speed, combined with its direct monetization model, amplify the damage. A 2025 Stanford University study estimated that duplicate content without attribution on X accounted for up to 12% of total engagement in the revenue program, diverting millions of dollars from original creators. The copy economy had become a parallel business: automated accounts downloaded videos from TikTok, Instagram Reels, or YouTube Shorts, reposted them on X, and earned revenue based on views from Premium subscribers, without generating original value.
Why is this important?
This measure is not just a matter of intellectual property; it is a change in X's business model. If original creators do not receive the economic benefit of their content, the incentive to keep creating on the platform disappears. By removing violators, X seeks to protect true content generators and maintain the long-term viability of its monetization ecosystem. Additionally, it demonstrates that artificial intelligence can be effectively applied to moderate content at scale, though it also raises questions about potential false positives and the transparency of the process. Historically, platforms have faced criticism for excessive or incorrect moderation: Facebook, for example, had to apologize in 2020 for mistakenly removing content about the Vietnam War. In X's case, reliance on Grok as a sole judge without a clear appeals process could breed distrust. However, detection speed is key: according to internal data leaked to The Verge in June 2026, Grok can now identify a duplicate video in less than 2 seconds after its publication, compared to 6-8 seconds for the previous version, drastically reducing exposure time and economic damage. This advance could mark a turning point in real-time content moderation.
Immediate and long-term consequences
For expelled creators, the loss of income can be significant, especially for those who relied on the program. According to estimates from Social Media Today, the 4,000 expelled creators represent approximately 2% of active participants in the revenue sharing program, but they accounted for up to 15% of duplicate traffic. This suggests the measure disproportionately impacts accounts with high volumes of non-original content. For the platform, the measure could increase trust among legitimate creators and improve content quality. A 2025 Bloomberg report already noted that the saturation of duplicate content on X was driving away original creators, who migrated to platforms like Bluesky or Threads. In the long term, this action could set a precedent for other social networks, which might adopt similar plagiarism detection systems. In fact, Meta is already testing an AI-based tool to detect reused content on Facebook and Instagram, according to Reuters. However, the risk of false positives is real: in 2024, YouTube mistakenly removed a video game creator's channel for using short clips of his own content, sparking a wave of criticism. X will need to implement a robust appeals system to avoid collateral damage. Additionally, the measure could incentivize creators to produce more original content or collaborate explicitly, strengthening the creative ecosystem.
What readers should know
- Detection is three times faster: Grok can now identify duplicate content before it accumulates massive engagement, reducing violators' exposure time. According to Bier, processing speed increased from 1,200 to 3,600 checks per second.
- Not just videos: Detection covers videos, images, and text, expanding the scope of moderation. It includes metadata comparison, audio fingerprints, and semantic text analysis.
- Expulsion from the revenue program: Expelled creators lose the ability to monetize their posts through the shared program, though they can still use the platform. It has not been specified whether they can reapply after a review period.
- Limited transparency: X has not detailed the exact criteria used by Grok or the appeals process, creating uncertainty among creators. Bier only mentioned that decisions are based on an undisclosed "similarity threshold."
"We don't know if the 4,000 expelled are all cases or just a first wave. The lack of transparency could breed distrust, but the direction is correct: protecting original content is vital for the health of any creator platform. The question is whether X can maintain the balance between automation and fairness, something other platforms have not yet fully achieved."